moses-smt/mosesdecoder
1
1
Fork 0
mirror of https://github.com/moses-smt/mosesdecoder.git synced 2024-10-26 11:28:48 +03:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' of https://github.com/moses-smt/mosesdecoder

Browse Source
This commit is contained in:
Philipp Koehn 2014-08-03 18:22:45 -04:00
commit 35c346378e
92 changed files with 1641 additions and 15540 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
OnDiskPt/TargetPhrase.cpp
View File

@ -257,7 +257,7 @@ Moses::TargetPhrase *TargetPhrase::ConvertToMoses(const std::vector<Moses::Facto
, const std::vector<float> &weightT
, bool isSyntax) const
{
Moses::TargetPhrase *ret = new Moses::TargetPhrase();
Moses::TargetPhrase *ret = new Moses::TargetPhrase(&phraseDict);
// words
size_t phraseSize = GetSize();

140
contrib/other-builds/extract-mixed-syntax/.cproject
View File

@ -1,140 +0,0 @@
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<?fileVersion 4.0.0?><cproject storage_type_id="org.eclipse.cdt.core.XmlProjectDescriptionStorage">
<storageModule moduleId="org.eclipse.cdt.core.settings">
<cconfiguration id="cdt.managedbuild.config.gnu.cross.exe.debug.1919499982">
<storageModule buildSystemId="org.eclipse.cdt.managedbuilder.core.configurationDataProvider" id="cdt.managedbuild.config.gnu.cross.exe.debug.1919499982" moduleId="org.eclipse.cdt.core.settings" name="Debug">
<externalSettings/>
<extensions>
<extension id="org.eclipse.cdt.core.GmakeErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.CWDLocator" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.GCCErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.GASErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.GLDErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.ELF" point="org.eclipse.cdt.core.BinaryParser"/>
</extensions>
</storageModule>
<storageModule moduleId="cdtBuildSystem" version="4.0.0">
<configuration artifactName="${ProjName}" buildArtefactType="org.eclipse.cdt.build.core.buildArtefactType.exe" buildProperties="org.eclipse.cdt.build.core.buildType=org.eclipse.cdt.build.core.buildType.debug,org.eclipse.cdt.build.core.buildArtefactType=org.eclipse.cdt.build.core.buildArtefactType.exe" cleanCommand="rm -rf" description="" id="cdt.managedbuild.config.gnu.cross.exe.debug.1919499982" name="Debug" parent="cdt.managedbuild.config.gnu.cross.exe.debug">
<folderInfo id="cdt.managedbuild.config.gnu.cross.exe.debug.1919499982." name="/" resourcePath="">
<toolChain id="cdt.managedbuild.toolchain.gnu.cross.exe.debug.456080129" name="Cross GCC" superClass="cdt.managedbuild.toolchain.gnu.cross.exe.debug">
<targetPlatform archList="all" binaryParser="org.eclipse.cdt.core.ELF" id="cdt.managedbuild.targetPlatform.gnu.cross.582801917" isAbstract="false" osList="all" superClass="cdt.managedbuild.targetPlatform.gnu.cross"/>
<builder buildPath="${workspace_loc:/extract-mixed-syntax/Debug}" id="cdt.managedbuild.builder.gnu.cross.1220166455" keepEnvironmentInBuildfile="false" managedBuildOn="true" name="Gnu Make Builder" parallelBuildOn="true" parallelizationNumber="optimal" superClass="cdt.managedbuild.builder.gnu.cross"/>
<tool id="cdt.managedbuild.tool.gnu.cross.c.compiler.1245611568" name="Cross GCC Compiler" superClass="cdt.managedbuild.tool.gnu.cross.c.compiler">
<option defaultValue="gnu.c.optimization.level.none" id="gnu.c.compiler.option.optimization.level.2055012191" name="Optimization Level" superClass="gnu.c.compiler.option.optimization.level" valueType="enumerated"/>
<option id="gnu.c.compiler.option.debugging.level.1768196213" name="Debug Level" superClass="gnu.c.compiler.option.debugging.level" value="gnu.c.debugging.level.max" valueType="enumerated"/>
<inputType id="cdt.managedbuild.tool.gnu.c.compiler.input.2007889843" superClass="cdt.managedbuild.tool.gnu.c.compiler.input"/>
</tool>
<tool id="cdt.managedbuild.tool.gnu.cross.cpp.compiler.1194558915" name="Cross G++ Compiler" superClass="cdt.managedbuild.tool.gnu.cross.cpp.compiler">
<option id="gnu.cpp.compiler.option.optimization.level.855436310" name="Optimization Level" superClass="gnu.cpp.compiler.option.optimization.level" value="gnu.cpp.compiler.optimization.level.none" valueType="enumerated"/>
<option id="gnu.cpp.compiler.option.debugging.level.506549229" name="Debug Level" superClass="gnu.cpp.compiler.option.debugging.level" value="gnu.cpp.compiler.debugging.level.max" valueType="enumerated"/>
<option id="gnu.cpp.compiler.option.include.paths.1497326561" name="Include paths (-I)" superClass="gnu.cpp.compiler.option.include.paths" valueType="includePath">
<listOptionValue builtIn="false" value="&quot;${workspace_loc}/../..&quot;"/>
<listOptionValue builtIn="false" value="&quot;${workspace_loc}/../../boost/include&quot;"/>
<listOptionValue builtIn="false" value="&quot;${workspace_loc:/${ProjName}}&quot;"/>
</option>
<inputType id="cdt.managedbuild.tool.gnu.cpp.compiler.input.2118510064" superClass="cdt.managedbuild.tool.gnu.cpp.compiler.input"/>
</tool>
<tool id="cdt.managedbuild.tool.gnu.cross.c.linker.606353571" name="Cross GCC Linker" superClass="cdt.managedbuild.tool.gnu.cross.c.linker"/>
<tool id="cdt.managedbuild.tool.gnu.cross.cpp.linker.740521305" name="Cross G++ Linker" superClass="cdt.managedbuild.tool.gnu.cross.cpp.linker">
<option id="gnu.cpp.link.option.libs.1946120010" name="Libraries (-l)" superClass="gnu.cpp.link.option.libs" valueType="libs">
<listOptionValue builtIn="false" value="boost_program_options-mt"/>
<listOptionValue builtIn="false" value="boost_iostreams-mt"/>
<listOptionValue builtIn="false" value="z"/>
</option>
<option id="gnu.cpp.link.option.paths.1563475751" name="Library search path (-L)" superClass="gnu.cpp.link.option.paths" valueType="libPaths">
<listOptionValue builtIn="false" value="&quot;${workspace_loc:}/../../boost/lib64&quot;"/>
</option>
<inputType id="cdt.managedbuild.tool.gnu.cpp.linker.input.106010037" superClass="cdt.managedbuild.tool.gnu.cpp.linker.input">
<additionalInput kind="additionalinputdependency" paths="$(USER_OBJS)"/>
<additionalInput kind="additionalinput" paths="$(LIBS)"/>
</inputType>
</tool>
<tool id="cdt.managedbuild.tool.gnu.cross.archiver.136661991" name="Cross GCC Archiver" superClass="cdt.managedbuild.tool.gnu.cross.archiver"/>
<tool id="cdt.managedbuild.tool.gnu.cross.assembler.2112208574" name="Cross GCC Assembler" superClass="cdt.managedbuild.tool.gnu.cross.assembler">
<inputType id="cdt.managedbuild.tool.gnu.assembler.input.172930211" superClass="cdt.managedbuild.tool.gnu.assembler.input"/>
</tool>
</toolChain>
</folderInfo>
<sourceEntries>
<entry excluding="util/sorted_uniform_test.cc|util/sized_iterator_test.cc|util/read_compressed_test.cc|util/probing_hash_table_test.cc|util/joint_sort_test.cc|util/multi_intersection_test.cc|util/file_piece_test.cc|util/bit_packing_test.cc" flags="VALUE_WORKSPACE_PATH|RESOLVED" kind="sourcePath" name=""/>
</sourceEntries>
</configuration>
</storageModule>
<storageModule moduleId="org.eclipse.cdt.core.externalSettings"/>
</cconfiguration>
<cconfiguration id="cdt.managedbuild.config.gnu.cross.exe.release.715007893">
<storageModule buildSystemId="org.eclipse.cdt.managedbuilder.core.configurationDataProvider" id="cdt.managedbuild.config.gnu.cross.exe.release.715007893" moduleId="org.eclipse.cdt.core.settings" name="Release">
<externalSettings/>
<extensions>
<extension id="org.eclipse.cdt.core.GmakeErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.CWDLocator" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.GCCErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.GASErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.GLDErrorParser" point="org.eclipse.cdt.core.ErrorParser"/>
<extension id="org.eclipse.cdt.core.ELF" point="org.eclipse.cdt.core.BinaryParser"/>
</extensions>
</storageModule>
<storageModule moduleId="cdtBuildSystem" version="4.0.0">
<configuration artifactName="${ProjName}" buildArtefactType="org.eclipse.cdt.build.core.buildArtefactType.exe" buildProperties="org.eclipse.cdt.build.core.buildType=org.eclipse.cdt.build.core.buildType.release,org.eclipse.cdt.build.core.buildArtefactType=org.eclipse.cdt.build.core.buildArtefactType.exe" cleanCommand="rm -rf" description="" id="cdt.managedbuild.config.gnu.cross.exe.release.715007893" name="Release" parent="cdt.managedbuild.config.gnu.cross.exe.release">
<folderInfo id="cdt.managedbuild.config.gnu.cross.exe.release.715007893." name="/" resourcePath="">
<toolChain id="cdt.managedbuild.toolchain.gnu.cross.exe.release.99436307" name="Cross GCC" superClass="cdt.managedbuild.toolchain.gnu.cross.exe.release">
<targetPlatform archList="all" binaryParser="org.eclipse.cdt.core.ELF" id="cdt.managedbuild.targetPlatform.gnu.cross.801178939" isAbstract="false" osList="all" superClass="cdt.managedbuild.targetPlatform.gnu.cross"/>
<builder buildPath="${workspace_loc:/extract-mixed-syntax/Release}" id="cdt.managedbuild.builder.gnu.cross.1999547547" keepEnvironmentInBuildfile="false" managedBuildOn="true" name="Gnu Make Builder" superClass="cdt.managedbuild.builder.gnu.cross"/>
<tool id="cdt.managedbuild.tool.gnu.cross.c.compiler.2138817906" name="Cross GCC Compiler" superClass="cdt.managedbuild.tool.gnu.cross.c.compiler">
<option defaultValue="gnu.c.optimization.level.most" id="gnu.c.compiler.option.optimization.level.1481537766" name="Optimization Level" superClass="gnu.c.compiler.option.optimization.level" valueType="enumerated"/>
<option id="gnu.c.compiler.option.debugging.level.1967527847" name="Debug Level" superClass="gnu.c.compiler.option.debugging.level" value="gnu.c.debugging.level.none" valueType="enumerated"/>
<inputType id="cdt.managedbuild.tool.gnu.c.compiler.input.442342681" superClass="cdt.managedbuild.tool.gnu.c.compiler.input"/>
</tool>
<tool id="cdt.managedbuild.tool.gnu.cross.cpp.compiler.1604862038" name="Cross G++ Compiler" superClass="cdt.managedbuild.tool.gnu.cross.cpp.compiler">
<option id="gnu.cpp.compiler.option.optimization.level.1847950300" name="Optimization Level" superClass="gnu.cpp.compiler.option.optimization.level" value="gnu.cpp.compiler.optimization.level.most" valueType="enumerated"/>
<option id="gnu.cpp.compiler.option.debugging.level.1130138972" name="Debug Level" superClass="gnu.cpp.compiler.option.debugging.level" value="gnu.cpp.compiler.debugging.level.none" valueType="enumerated"/>
<inputType id="cdt.managedbuild.tool.gnu.cpp.compiler.input.870650754" superClass="cdt.managedbuild.tool.gnu.cpp.compiler.input"/>
</tool>
<tool id="cdt.managedbuild.tool.gnu.cross.c.linker.158429528" name="Cross GCC Linker" superClass="cdt.managedbuild.tool.gnu.cross.c.linker"/>
<tool id="cdt.managedbuild.tool.gnu.cross.cpp.linker.2020667840" name="Cross G++ Linker" superClass="cdt.managedbuild.tool.gnu.cross.cpp.linker">
<inputType id="cdt.managedbuild.tool.gnu.cpp.linker.input.1372779734" superClass="cdt.managedbuild.tool.gnu.cpp.linker.input">
<additionalInput kind="additionalinputdependency" paths="$(USER_OBJS)"/>
<additionalInput kind="additionalinput" paths="$(LIBS)"/>
</inputType>
</tool>
<tool id="cdt.managedbuild.tool.gnu.cross.archiver.371006952" name="Cross GCC Archiver" superClass="cdt.managedbuild.tool.gnu.cross.archiver"/>
<tool id="cdt.managedbuild.tool.gnu.cross.assembler.1770045040" name="Cross GCC Assembler" superClass="cdt.managedbuild.tool.gnu.cross.assembler">
<inputType id="cdt.managedbuild.tool.gnu.assembler.input.707592414" superClass="cdt.managedbuild.tool.gnu.assembler.input"/>
</tool>
</toolChain>
</folderInfo>
</configuration>
</storageModule>
<storageModule moduleId="org.eclipse.cdt.core.externalSettings"/>
</cconfiguration>
</storageModule>
<storageModule moduleId="cdtBuildSystem" version="4.0.0">
<project id="extract-mixed-syntax.cdt.managedbuild.target.gnu.cross.exe.1868010260" name="Executable" projectType="cdt.managedbuild.target.gnu.cross.exe"/>
</storageModule>
<storageModule moduleId="scannerConfiguration">
<autodiscovery enabled="true" problemReportingEnabled="true" selectedProfileId=""/>
<scannerConfigBuildInfo instanceId="cdt.managedbuild.config.gnu.cross.exe.release.715007893;cdt.managedbuild.config.gnu.cross.exe.release.715007893.;cdt.managedbuild.tool.gnu.cross.cpp.compiler.1604862038;cdt.managedbuild.tool.gnu.cpp.compiler.input.870650754">
<autodiscovery enabled="true" problemReportingEnabled="true" selectedProfileId="org.eclipse.cdt.managedbuilder.core.GCCManagedMakePerProjectProfileCPP"/>
</scannerConfigBuildInfo>
<scannerConfigBuildInfo instanceId="cdt.managedbuild.config.gnu.cross.exe.release.715007893;cdt.managedbuild.config.gnu.cross.exe.release.715007893.;cdt.managedbuild.tool.gnu.cross.c.compiler.2138817906;cdt.managedbuild.tool.gnu.c.compiler.input.442342681">
<autodiscovery enabled="true" problemReportingEnabled="true" selectedProfileId="org.eclipse.cdt.managedbuilder.core.GCCManagedMakePerProjectProfileC"/>
</scannerConfigBuildInfo>
<scannerConfigBuildInfo instanceId="cdt.managedbuild.config.gnu.cross.exe.debug.1919499982;cdt.managedbuild.config.gnu.cross.exe.debug.1919499982.;cdt.managedbuild.tool.gnu.cross.cpp.compiler.1194558915;cdt.managedbuild.tool.gnu.cpp.compiler.input.2118510064">
<autodiscovery enabled="true" problemReportingEnabled="true" selectedProfileId="org.eclipse.cdt.managedbuilder.core.GCCManagedMakePerProjectProfileCPP"/>
</scannerConfigBuildInfo>
<scannerConfigBuildInfo instanceId="cdt.managedbuild.config.gnu.cross.exe.debug.1919499982;cdt.managedbuild.config.gnu.cross.exe.debug.1919499982.;cdt.managedbuild.tool.gnu.cross.c.compiler.1245611568;cdt.managedbuild.tool.gnu.c.compiler.input.2007889843">
<autodiscovery enabled="true" problemReportingEnabled="true" selectedProfileId="org.eclipse.cdt.managedbuilder.core.GCCManagedMakePerProjectProfileC"/>
</scannerConfigBuildInfo>
</storageModule>
<storageModule moduleId="org.eclipse.cdt.core.LanguageSettingsProviders"/>
<storageModule moduleId="refreshScope" versionNumber="2">
<configuration configurationName="Release">
<resource resourceType="PROJECT" workspacePath="/extract-mixed-syntax"/>
</configuration>
<configuration configurationName="Debug">
<resource resourceType="PROJECT" workspacePath="/extract-mixed-syntax"/>
</configuration>
</storageModule>
<storageModule moduleId="org.eclipse.cdt.internal.ui.text.commentOwnerProjectMappings"/>
<storageModule moduleId="org.eclipse.cdt.make.core.buildtargets"/>
</cproject>

27
contrib/other-builds/extract-mixed-syntax/.project
View File

@ -1,27 +0,0 @@
<?xml version="1.0" encoding="UTF-8"?>
<projectDescription>
<name>extract-mixed-syntax</name>
<comment></comment>
<projects>
</projects>
<buildSpec>
<buildCommand>
<name>org.eclipse.cdt.managedbuilder.core.genmakebuilder</name>
<triggers>clean,full,incremental,</triggers>
<arguments>
</arguments>
</buildCommand>
<buildCommand>
<name>org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder</name>
<triggers>full,incremental,</triggers>
<arguments>
</arguments>
</buildCommand>
</buildSpec>
<natures>
<nature>org.eclipse.cdt.core.cnature</nature>
<nature>org.eclipse.cdt.core.ccnature</nature>
<nature>org.eclipse.cdt.managedbuilder.core.managedBuildNature</nature>
<nature>org.eclipse.cdt.managedbuilder.core.ScannerConfigNature</nature>
</natures>
</projectDescription>

189
contrib/other-builds/extract-mixed-syntax/AlignedSentence.cpp
View File

@ -1,189 +0,0 @@
/*
* AlignedSentence.cpp
*
* Created on: 18 Feb 2014
* Author: s0565741
*/
#include <sstream>
#include "moses/Util.h"
#include "AlignedSentence.h"
#include "Parameter.h"
using namespace std;
/////////////////////////////////////////////////////////////////////////////////
AlignedSentence::AlignedSentence(int lineNum,
const std::string &source,
const std::string &target,
const std::string &alignment)
:m_lineNum(lineNum)
{
PopulateWordVec(m_source, source);
PopulateWordVec(m_target, target);
PopulateAlignment(alignment);
}
AlignedSentence::~AlignedSentence() {
Moses::RemoveAllInColl(m_source);
Moses::RemoveAllInColl(m_target);
}
void AlignedSentence::PopulateWordVec(Phrase &vec, const std::string &line)
{
std::vector<string> toks;
Moses::Tokenize(toks, line);
vec.resize(toks.size());
for (size_t i = 0; i < vec.size(); ++i) {
const string &tok = toks[i];
Word *word = new Word(i, tok);
vec[i] = word;
}
}
void AlignedSentence::PopulateAlignment(const std::string &line)
{
vector<string> alignStr;
Moses::Tokenize(alignStr, line);
for (size_t i = 0; i < alignStr.size(); ++i) {
vector<int> alignPair;
Moses::Tokenize(alignPair, alignStr[i], "-");
assert(alignPair.size() == 2);
int sourcePos = alignPair[0];
int targetPos = alignPair[1];
if (sourcePos >= m_source.size()) {
cerr << "ERROR1:AlignedSentence=" << Debug() << endl;
cerr << "m_source=" << m_source.size() << endl;
abort();
}
assert(sourcePos < m_source.size());
assert(targetPos < m_target.size());
Word *sourceWord = m_source[sourcePos];
Word *targetWord = m_target[targetPos];
sourceWord->AddAlignment(targetWord);
targetWord->AddAlignment(sourceWord);
}
}
std::string AlignedSentence::Debug() const
{
stringstream out;
out << "m_lineNum:";
out << m_lineNum;
out << endl;
out << "m_source:";
out << m_source.Debug();
out << endl;
out << "m_target:";
out << m_target.Debug();
out << endl;
out << "consistent phrases:" << endl;
out << m_consistentPhrases.Debug();
out << endl;
return out.str();
}
std::vector<int> AlignedSentence::GetSourceAlignmentCount() const
{
vector<int> ret(m_source.size());
for (size_t i = 0; i < m_source.size(); ++i) {
const Word &word = *m_source[i];
ret[i] = word.GetAlignmentIndex().size();
}
return ret;
}
void AlignedSentence::Create(const Parameter &params)
{
CreateConsistentPhrases(params);
m_consistentPhrases.AddHieroNonTerms(params);
}
void AlignedSentence::CreateConsistentPhrases(const Parameter &params)
{
int countT = m_target.size();
int countS = m_source.size();
m_consistentPhrases.Initialize(countS);
// check alignments for target phrase startT...endT
for(int lengthT=1;
lengthT <= params.maxSpan && lengthT <= countT;
lengthT++) {
for(int startT=0; startT < countT-(lengthT-1); startT++) {
// that's nice to have
int endT = startT + lengthT - 1;
// find find aligned source words
// first: find minimum and maximum source word
int minS = 9999;
int maxS = -1;
vector< int > usedS = GetSourceAlignmentCount();
for(int ti=startT; ti<=endT; ti++) {
const Word &word = *m_target[ti];
const std::set<int> &alignment = word.GetAlignmentIndex();
std::set<int>::const_iterator iterAlign;
for(iterAlign = alignment.begin(); iterAlign != alignment.end(); ++iterAlign) {
int si = *iterAlign;
if (si<minS) {
minS = si;
}
if (si>maxS) {
maxS = si;
}
usedS[ si ]--;
}
}
// unaligned phrases are not allowed
if( maxS == -1 )
continue;
// source phrase has to be within limits
if( maxS-minS >= params.maxSpan )
continue;
// check if source words are aligned to out of bound target words
bool out_of_bounds = false;
for(int si=minS; si<=maxS && !out_of_bounds; si++)
if (usedS[si]>0) {
out_of_bounds = true;
}
// if out of bound, you gotta go
if (out_of_bounds)
continue;
// done with all the checks, lets go over all consistent phrase pairs
// start point of source phrase may retreat over unaligned
for(int startS=minS;
(startS>=0 &&
startS>maxS - params.maxSpan && // within length limit
(startS==minS || m_source[startS]->GetAlignment().size()==0)); // unaligned
startS--) {
// end point of source phrase may advance over unaligned
for(int endS=maxS;
(endS<countS && endS<startS + params.maxSpan && // within length limit
(endS==maxS || m_source[endS]->GetAlignment().size()==0)); // unaligned
endS++) {
// take note that this is a valid phrase alignment
m_consistentPhrases.Add(startS, endS, startT, endT, params);
}
}
}
}
}

51
contrib/other-builds/extract-mixed-syntax/AlignedSentence.h
View File

@ -1,51 +0,0 @@
/*
* AlignedSentence.h
*
* Created on: 18 Feb 2014
* Author: s0565741
*/
#pragma once
#include <string>
#include <set>
#include "ConsistentPhrases.h"
#include "Phrase.h"
#include "moses/TypeDef.h"
class Parameter;
class AlignedSentence {
public:
AlignedSentence(int lineNum)
:m_lineNum(lineNum)
{}
AlignedSentence(int lineNum,
const std::string &source,
const std::string &target,
const std::string &alignment);
virtual ~AlignedSentence();
virtual void Create(const Parameter &params);
const Phrase &GetPhrase(Moses::FactorDirection direction) const
{ return (direction == Moses::Input) ? m_source : m_target; }
const ConsistentPhrases &GetConsistentPhrases() const
{ return m_consistentPhrases; }
virtual std::string Debug() const;
int m_lineNum;
protected:
Phrase m_source, m_target;
ConsistentPhrases m_consistentPhrases;
void CreateConsistentPhrases(const Parameter &params);
void PopulateWordVec(Phrase &vec, const std::string &line);
// m_source and m_target MUST be populated before calling this
void PopulateAlignment(const std::string &line);
std::vector<int> GetSourceAlignmentCount() const;
};

183
contrib/other-builds/extract-mixed-syntax/AlignedSentenceSyntax.cpp
View File

@ -1,183 +0,0 @@
/*
* AlignedSentenceSyntax.cpp
*
* Created on: 26 Feb 2014
* Author: hieu
*/
#include "AlignedSentenceSyntax.h"
#include "Parameter.h"
#include "pugixml.hpp"
#include "moses/Util.h"
using namespace std;
AlignedSentenceSyntax::AlignedSentenceSyntax(int lineNum,
const std::string &source,
const std::string &target,
const std::string &alignment)
:AlignedSentence(lineNum)
,m_sourceStr(source)
,m_targetStr(target)
,m_alignmentStr(alignment)
{
}
AlignedSentenceSyntax::~AlignedSentenceSyntax() {
// TODO Auto-generated destructor stub
}
void AlignedSentenceSyntax::Populate(bool isSyntax, int mixedSyntaxType, const Parameter &params,
string line, Phrase &phrase, SyntaxTree &tree)
{
// parse source and target string
if (isSyntax) {
line = "<xml><tree label=\"X\">" + line + "</tree></xml>";
XMLParse(phrase, tree, line, params);
if (mixedSyntaxType != 0) {
// mixed syntax. Always add [X] where there isn't 1
tree.SetHieroLabel(params.hieroNonTerm);
if (mixedSyntaxType == 2) {
tree.AddToAll(params.hieroNonTerm);
}
}
}
else {
PopulateWordVec(phrase, line);
tree.SetHieroLabel(params.hieroNonTerm);
}
}
void AlignedSentenceSyntax::Create(const Parameter &params)
{
Populate(params.sourceSyntax, params.mixedSyntaxType, params, m_sourceStr,
m_source, m_sourceTree);
Populate(params.targetSyntax, params.mixedSyntaxType, params, m_targetStr,
m_target, m_targetTree);
PopulateAlignment(m_alignmentStr);
CreateConsistentPhrases(params);
// create labels
CreateNonTerms();
}
void Escape(string &text)
{
text = Moses::Replace(text, "&", "&amp;");
text = Moses::Replace(text, "|", "&#124;");
text = Moses::Replace(text, "<", "&lt;");
text = Moses::Replace(text, ">", "&gt;");
text = Moses::Replace(text, "'", "&apos;");
text = Moses::Replace(text, "\"", "&quot;");
text = Moses::Replace(text, "[", "&#91;");
text = Moses::Replace(text, "]", "&#93;");
}
void AlignedSentenceSyntax::XMLParse(Phrase &output,
SyntaxTree &tree,
const pugi::xml_node &parentNode,
const Parameter &params)
{
int childNum = 0;
for (pugi::xml_node childNode = parentNode.first_child(); childNode; childNode = childNode.next_sibling())
{
string nodeName = childNode.name();
// span label
string label;
int startPos = output.size();
if (!nodeName.empty()) {
pugi::xml_attribute attribute = childNode.attribute("label");
label = attribute.as_string();
// recursively call this function. For proper recursive trees
XMLParse(output, tree, childNode, params);
}
// fill phrase vector
string text = childNode.value();
Escape(text);
//cerr << childNum << " " << label << "=" << text << endl;
std::vector<string> toks;
Moses::Tokenize(toks, text);
for (size_t i = 0; i < toks.size(); ++i) {
const string &tok = toks[i];
Word *word = new Word(output.size(), tok);
output.push_back(word);
}
// is it a labelled span?
int endPos = output.size() - 1;
// fill syntax labels
if (!label.empty()) {
label = "[" + label + "]";
tree.Add(startPos, endPos, label, params);
}
++childNum;
}
}
void AlignedSentenceSyntax::XMLParse(Phrase &output,
SyntaxTree &tree,
const std::string input,
const Parameter &params)
{
pugi::xml_document doc;
pugi::xml_parse_result result = doc.load(input.c_str(),
pugi::parse_default | pugi::parse_comments);
pugi::xml_node topNode = doc.child("xml");
XMLParse(output, tree, topNode, params);
}
void AlignedSentenceSyntax::CreateNonTerms()
{
for (int sourceStart = 0; sourceStart < m_source.size(); ++sourceStart) {
for (int sourceEnd = sourceStart; sourceEnd < m_source.size(); ++sourceEnd) {
ConsistentPhrases::Coll &coll = m_consistentPhrases.GetColl(sourceStart, sourceEnd);
const SyntaxTree::Labels &sourceLabels = m_sourceTree.Find(sourceStart, sourceEnd);
ConsistentPhrases::Coll::iterator iter;
for (iter = coll.begin(); iter != coll.end(); ++iter) {
ConsistentPhrase &cp = **iter;
int targetStart = cp.corners[2];
int targetEnd = cp.corners[3];
const SyntaxTree::Labels &targetLabels = m_targetTree.Find(targetStart, targetEnd);
CreateNonTerms(cp, sourceLabels, targetLabels);
}
}
}
}
void AlignedSentenceSyntax::CreateNonTerms(ConsistentPhrase &cp,
const SyntaxTree::Labels &sourceLabels,
const SyntaxTree::Labels &targetLabels)
{
SyntaxTree::Labels::const_iterator iterSource;
for (iterSource = sourceLabels.begin(); iterSource != sourceLabels.end(); ++iterSource) {
const string &sourceLabel = *iterSource;
SyntaxTree::Labels::const_iterator iterTarget;
for (iterTarget = targetLabels.begin(); iterTarget != targetLabels.end(); ++iterTarget) {
const string &targetLabel = *iterTarget;
cp.AddNonTerms(sourceLabel, targetLabel);
}
}
}

46
contrib/other-builds/extract-mixed-syntax/AlignedSentenceSyntax.h
View File

@ -1,46 +0,0 @@
/*
* AlignedSentenceSyntax.h
*
* Created on: 26 Feb 2014
* Author: hieu
*/
#pragma once
#include "AlignedSentence.h"
#include "SyntaxTree.h"
#include "pugixml.hpp"
class AlignedSentenceSyntax : public AlignedSentence
{
public:
AlignedSentenceSyntax(int lineNum,
const std::string &source,
const std::string &target,
const std::string &alignment);
virtual ~AlignedSentenceSyntax();
void Create(const Parameter &params);
//virtual std::string Debug() const;
protected:
std::string m_sourceStr, m_targetStr, m_alignmentStr;
SyntaxTree m_sourceTree, m_targetTree;
void XMLParse(Phrase &output,
SyntaxTree &tree,
const std::string input,
const Parameter &params);
void XMLParse(Phrase &output,
SyntaxTree &tree,
const pugi::xml_node &parentNode,
const Parameter &params);
void CreateNonTerms();
void CreateNonTerms(ConsistentPhrase &cp,
const SyntaxTree::Labels &sourceLabels,
const SyntaxTree::Labels &targetLabels);
void Populate(bool isSyntax, int mixedSyntaxType, const Parameter &params,
std::string line, Phrase &phrase, SyntaxTree &tree);
};

66
contrib/other-builds/extract-mixed-syntax/ConsistentPhrase.cpp
View File

@ -1,66 +0,0 @@
/*
* ConsistentPhrase.cpp
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#include <sstream>
#include "ConsistentPhrase.h"
#include "Word.h"
#include "NonTerm.h"
#include "Parameter.h"
using namespace std;
ConsistentPhrase::ConsistentPhrase(
int sourceStart, int sourceEnd,
int targetStart, int targetEnd,
const Parameter &params)
:corners(4)
,m_hieroNonTerm(*this, params.hieroNonTerm, params.hieroNonTerm)
{
corners[0] = sourceStart;
corners[1] = sourceEnd;
corners[2] = targetStart;
corners[3] = targetEnd;
}
ConsistentPhrase::~ConsistentPhrase() {
// TODO Auto-generated destructor stub
}
bool ConsistentPhrase::operator<(const ConsistentPhrase &other) const
{
return corners < other.corners;
}
void ConsistentPhrase::AddNonTerms(const std::string &source,
const std::string &target)
{
m_nonTerms.push_back(NonTerm(*this, source, target));
}
bool ConsistentPhrase::TargetOverlap(const ConsistentPhrase &other) const
{
if ( other.corners[3] < corners[2] || other.corners[2] > corners[3])
return false;
return true;
}
std::string ConsistentPhrase::Debug() const
{
stringstream out;
out << "[" << corners[0] << "-" << corners[1]
<< "][" << corners[2] << "-" << corners[3] << "]";
out << "NT:";
for (size_t i = 0; i < m_nonTerms.size(); ++i) {
const NonTerm &nonTerm = m_nonTerms[i];
out << nonTerm.GetLabel(Moses::Input) << ":" << nonTerm.GetLabel(Moses::Output);
}
return out.str();
}

51
contrib/other-builds/extract-mixed-syntax/ConsistentPhrase.h
View File

@ -1,51 +0,0 @@
/*
* ConsistentPhrase.h
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#pragma once
#include <cassert>
#include <vector>
#include <iostream>
#include "moses/TypeDef.h"
#include "NonTerm.h"
class ConsistentPhrase
{
public:
typedef std::vector<NonTerm> NonTerms;
std::vector<int> corners;
ConsistentPhrase(const ConsistentPhrase &copy); // do not implement
ConsistentPhrase(int sourceStart, int sourceEnd,
int targetStart, int targetEnd,
const Parameter &params);
virtual ~ConsistentPhrase();
int GetWidth(Moses::FactorDirection direction) const
{ return (direction == Moses::Input) ? corners[1] - corners[0] + 1 : corners[3] - corners[2] + 1; }
void AddNonTerms(const std::string &source,
const std::string &target);
const NonTerms &GetNonTerms() const
{ return m_nonTerms;}
const NonTerm &GetHieroNonTerm() const
{ return m_hieroNonTerm;}
bool TargetOverlap(const ConsistentPhrase &other) const;
bool operator<(const ConsistentPhrase &other) const;
std::string Debug() const;
protected:
NonTerms m_nonTerms;
NonTerm m_hieroNonTerm;
};

103
contrib/other-builds/extract-mixed-syntax/ConsistentPhrases.cpp
View File

@ -1,103 +0,0 @@
/*
* ConsistentPhrases.cpp
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#include <sstream>
#include <cassert>
#include "ConsistentPhrases.h"
#include "NonTerm.h"
#include "Parameter.h"
#include "moses/Util.h"
using namespace std;
ConsistentPhrases::ConsistentPhrases()
{
}
ConsistentPhrases::~ConsistentPhrases() {
for (int start = 0; start < m_coll.size(); ++start) {
std::vector<Coll> &allSourceStart = m_coll[start];
for (int size = 0; size < allSourceStart.size(); ++size) {
Coll &coll = allSourceStart[size];
Moses::RemoveAllInColl(coll);
}
}
}
void ConsistentPhrases::Initialize(size_t size)
{
m_coll.resize(size);
for (size_t sourceStart = 0; sourceStart < size; ++sourceStart) {
std::vector<Coll> &allSourceStart = m_coll[sourceStart];
allSourceStart.resize(size - sourceStart);
}
}
void ConsistentPhrases::Add(int sourceStart, int sourceEnd,
int targetStart, int targetEnd,
const Parameter &params)
{
Coll &coll = m_coll[sourceStart][sourceEnd - sourceStart];
ConsistentPhrase *cp = new ConsistentPhrase(sourceStart, sourceEnd,
targetStart, targetEnd,
params);
pair<Coll::iterator, bool> inserted = coll.insert(cp);
assert(inserted.second);
}
const ConsistentPhrases::Coll &ConsistentPhrases::GetColl(int sourceStart, int sourceEnd) const
{
const std::vector<Coll> &allSourceStart = m_coll[sourceStart];
const Coll &ret = allSourceStart[sourceEnd - sourceStart];
return ret;
}
ConsistentPhrases::Coll &ConsistentPhrases::GetColl(int sourceStart, int sourceEnd)
{
std::vector<Coll> &allSourceStart = m_coll[sourceStart];
Coll &ret = allSourceStart[sourceEnd - sourceStart];
return ret;
}
std::string ConsistentPhrases::Debug() const
{
std::stringstream out;
for (int start = 0; start < m_coll.size(); ++start) {
const std::vector<Coll> &allSourceStart = m_coll[start];
for (int size = 0; size < allSourceStart.size(); ++size) {
const Coll &coll = allSourceStart[size];
Coll::const_iterator iter;
for (iter = coll.begin(); iter != coll.end(); ++iter) {
const ConsistentPhrase &consistentPhrase = **iter;
out << consistentPhrase.Debug() << endl;
}
}
}
return out.str();
}
void ConsistentPhrases::AddHieroNonTerms(const Parameter &params)
{
// add [X] labels everywhere
for (int i = 0; i < m_coll.size(); ++i) {
vector<Coll> &inner = m_coll[i];
for (int j = 0; j < inner.size(); ++j) {
ConsistentPhrases::Coll &coll = inner[j];
ConsistentPhrases::Coll::iterator iter;
for (iter = coll.begin(); iter != coll.end(); ++iter) {
ConsistentPhrase &cp = **iter;
cp.AddNonTerms(params.hieroNonTerm, params.hieroNonTerm);
}
}
}
}

40
contrib/other-builds/extract-mixed-syntax/ConsistentPhrases.h
View File

@ -1,40 +0,0 @@
/*
* ConsistentPhrases.h
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#pragma once
#include <set>
#include <vector>
#include <iostream>
#include "ConsistentPhrase.h"
class Word;
class Parameter;
class ConsistentPhrases {
public:
typedef std::set<ConsistentPhrase*> Coll;
ConsistentPhrases();
virtual ~ConsistentPhrases();
void Initialize(size_t size);
void Add(int sourceStart, int sourceEnd,
int targetStart, int targetEnd,
const Parameter &params);
void AddHieroNonTerms(const Parameter &params);
const Coll &GetColl(int sourceStart, int sourceEnd) const;
Coll &GetColl(int sourceStart, int sourceEnd);
std::string Debug() const;
protected:
std::vector< std::vector<Coll> > m_coll;
};

62
contrib/other-builds/extract-mixed-syntax/InputFileStream.cpp
View File

@ -1,62 +0,0 @@
// $Id: InputFileStream.cpp 2780 2010-01-29 17:11:17Z bojar $
/***********************************************************************
Moses - factored phrase-based language decoder
Copyright (C) 2006 University of Edinburgh
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
***********************************************************************/
#include "InputFileStream.h"
#include "gzfilebuf.h"
#include <iostream>
using namespace std;
namespace Moses
{
InputFileStream::InputFileStream(const std::string &filePath)
: std::istream(NULL)
, m_streambuf(NULL)
{
if (filePath.size() > 3 &&
filePath.substr(filePath.size() - 3, 3) == ".gz")
{
m_streambuf = new gzfilebuf(filePath.c_str());
} else {
std::filebuf* fb = new std::filebuf();
fb = fb->open(filePath.c_str(), std::ios::in);
if (! fb) {
cerr << "Can't read " << filePath.c_str() << endl;
exit(1);
}
m_streambuf = fb;
}
this->init(m_streambuf);
}
InputFileStream::~InputFileStream()
{
delete m_streambuf;
m_streambuf = NULL;
}
void InputFileStream::Close()
{
}
}

48
contrib/other-builds/extract-mixed-syntax/InputFileStream.h
View File

@ -1,48 +0,0 @@
// $Id: InputFileStream.h 2939 2010-02-24 11:15:44Z jfouet $
/***********************************************************************
Moses - factored phrase-based language decoder
Copyright (C) 2006 University of Edinburgh
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
***********************************************************************/
#ifndef moses_InputFileStream_h
#define moses_InputFileStream_h
#include <cstdlib>
#include <fstream>
#include <string>
namespace Moses
{
/** Used in place of std::istream, can read zipped files if it ends in .gz
*/
class InputFileStream : public std::istream
{
protected:
std::streambuf *m_streambuf;
public:
InputFileStream(const std::string &filePath);
~InputFileStream();
void Close();
};
}
#endif

174
contrib/other-builds/extract-mixed-syntax/Main.cpp
View File

@ -1,174 +0,0 @@
#include <iostream>
#include <cstdlib>
#include <boost/program_options.hpp>
#include "Main.h"
#include "InputFileStream.h"
#include "OutputFileStream.h"
#include "AlignedSentence.h"
#include "AlignedSentenceSyntax.h"
#include "Parameter.h"
#include "Rules.h"
using namespace std;
bool g_debug = false;
int main(int argc, char** argv)
{
cerr << "Starting" << endl;
Parameter params;
namespace po = boost::program_options;
po::options_description desc("Options");
desc.add_options()
("help", "Print help messages")
("MaxSpan", po::value<int>()->default_value(params.maxSpan), "Max (source) span of a rule. ie. number of words in the source")
("GlueGrammar", po::value<string>()->default_value(params.gluePath), "Output glue grammar to here")
("SentenceOffset", po::value<long>()->default_value(params.sentenceOffset), "Starting sentence id. Not used")
("GZOutput", "Compress extract files")
("MaxNonTerm", po::value<int>()->default_value(params.maxNonTerm), "Maximum number of non-terms allowed per rule")
("MaxHieroNonTerm", po::value<int>()->default_value(params.maxHieroNonTerm), "Maximum number of Hiero non-term. Usually, --MaxNonTerm is the normal constraint")
("MinHoleSource", po::value<int>()->default_value(params.minHoleSource), "Minimum source span for a non-term.")
("SourceSyntax", "Source sentence is a parse tree")
("TargetSyntax", "Target sentence is a parse tree")
("MixedSyntaxType", po::value<int>()->default_value(params.mixedSyntaxType), "Hieu's Mixed syntax type. 0(default)=no mixed syntax, 1=add [X] only if no syntactic label. 2=add [X] everywhere")
("MultiLabel", po::value<int>()->default_value(params.multiLabel), "What to do with multiple labels on the same span. 0(default)=keep them all, 1=keep only top-most, 2=keep only bottom-most")
("HieroSourceLHS", "Always use Hiero source LHS? Default = 0")
("MaxSpanFreeNonTermSource", po::value<int>()->default_value(params.maxSpanFreeNonTermSource), "Max number of words covered by beginning/end NT. Default = 0 (no limit)")
("NoNieceTerminal", "Don't extract rule if 1 of the non-term covers the same word as 1 of the terminals")
("MaxScope", po::value<int>()->default_value(params.maxScope), "maximum scope (see Hopkins and Langmead (2010)). Default is HIGH")
("SpanLength", "Property - span length of RHS each non-term")
("NonTermContext", "Property - left and right, inside and outside words of each non-term");
po::variables_map vm;
try
{
po::store(po::parse_command_line(argc, argv, desc),
vm); // can throw
/** --help option
*/
if ( vm.count("help") || argc < 5 )
{
std::cout << argv[0] << " target source alignment [options...]" << std::endl
<< desc << std::endl;
return EXIT_SUCCESS;
}
po::notify(vm); // throws on error, so do after help in case
// there are any problems
}
catch(po::error& e)
{
std::cerr << "ERROR: " << e.what() << std::endl << std::endl;
std::cerr << desc << std::endl;
return EXIT_FAILURE;
}
if (vm.count("MaxSpan")) params.maxSpan = vm["MaxSpan"].as<int>();
if (vm.count("GZOutput")) params.gzOutput = true;
if (vm.count("GlueGrammar")) params.gluePath = vm["GlueGrammar"].as<string>();
if (vm.count("SentenceOffset")) params.sentenceOffset = vm["SentenceOffset"].as<long>();
if (vm.count("MaxNonTerm")) params.maxNonTerm = vm["MaxNonTerm"].as<int>();
if (vm.count("MaxHieroNonTerm")) params.maxHieroNonTerm = vm["MaxHieroNonTerm"].as<int>();
if (vm.count("MinHoleSource")) params.minHoleSource = vm["MinHoleSource"].as<int>();
if (vm.count("SourceSyntax")) params.sourceSyntax = true;
if (vm.count("TargetSyntax")) params.targetSyntax = true;
if (vm.count("MixedSyntaxType")) params.mixedSyntaxType = vm["MixedSyntaxType"].as<int>();
if (vm.count("MultiLabel")) params.multiLabel = vm["MultiLabel"].as<int>();
if (vm.count("HieroSourceLHS")) params.hieroSourceLHS = true;
if (vm.count("MaxSpanFreeNonTermSource")) params.maxSpanFreeNonTermSource = vm["MaxSpanFreeNonTermSource"].as<int>();
if (vm.count("NoNieceTerminal")) params.nieceTerminal = false;
if (vm.count("MaxScope")) params.maxScope = vm["MaxScope"].as<int>();
// properties
if (vm.count("SpanLength")) params.spanLength = true;
if (vm.count("NonTermContext")) params.nonTermContext = true;
// input files;
string pathTarget = argv[1];
string pathSource = argv[2];
string pathAlignment = argv[3];
string pathExtract = argv[4];
string pathExtractInv = pathExtract + ".inv";
if (params.gzOutput) {
pathExtract += ".gz";
pathExtractInv += ".gz";
}
Moses::InputFileStream strmTarget(pathTarget);
Moses::InputFileStream strmSource(pathSource);
Moses::InputFileStream strmAlignment(pathAlignment);
Moses::OutputFileStream extractFile(pathExtract);
Moses::OutputFileStream extractInvFile(pathExtractInv);
// MAIN LOOP
int lineNum = 1;
string lineTarget, lineSource, lineAlignment;
while (getline(strmTarget, lineTarget)) {
if (lineNum % 10000 == 0) {
cerr << lineNum << " ";
}
bool success;
success = getline(strmSource, lineSource);
if (!success) {
throw "Couldn't read source";
}
success = getline(strmAlignment, lineAlignment);
if (!success) {
throw "Couldn't read alignment";
}
/*
cerr << "lineTarget=" << lineTarget << endl;
cerr << "lineSource=" << lineSource << endl;
cerr << "lineAlignment=" << lineAlignment << endl;
*/
AlignedSentence *alignedSentence;
if (params.sourceSyntax || params.targetSyntax) {
alignedSentence = new AlignedSentenceSyntax(lineNum, lineSource, lineTarget, lineAlignment);
}
else {
alignedSentence = new AlignedSentence(lineNum, lineSource, lineTarget, lineAlignment);
}
alignedSentence->Create(params);
//cerr << alignedSentence->Debug();
Rules rules(*alignedSentence);
rules.Extend(params);
rules.Consolidate(params);
//cerr << rules.Debug();
rules.Output(extractFile, true, params);
rules.Output(extractInvFile, false, params);
delete alignedSentence;
++lineNum;
}
if (!params.gluePath.empty()) {
Moses::OutputFileStream glueFile(params.gluePath);
CreateGlueGrammar(glueFile);
}
cerr << "Finished" << endl;
}
void CreateGlueGrammar(Moses::OutputFileStream &glueFile)
{
glueFile << "<s> [X] ||| <s> [S] ||| 1 ||| ||| 0" << endl
<< "[X][S] </s> [X] ||| [X][S] </s> [S] ||| 1 ||| 0-0 ||| 0" << endl
<< "[X][S] [X][X] [X] ||| [X][S] [X][X] [S] ||| 2.718 ||| 0-0 1-1 ||| 0" << endl;
}

12
contrib/other-builds/extract-mixed-syntax/Main.h
View File

@ -1,12 +0,0 @@
/*
* Main.h
*
* Created on: 28 Feb 2014
* Author: hieu
*/
#pragma once
#include "OutputFileStream.h"
void CreateGlueGrammar(Moses::OutputFileStream &glueFile);

17
contrib/other-builds/extract-mixed-syntax/Makefile
View File

@ -1,17 +0,0 @@
all: extract-mixed-syntax
clean:
rm -f *.o extract-mixed-syntax
.cpp.o:
g++ -O4 -g -c -I../../../boost/include -I../../../ $<
OBJECTS = AlignedSentence.o ConsistentPhrase.o ConsistentPhrases.o InputFileStream.o \
Main.o OutputFileStream.o Parameter.o Phrase.o Rule.o Rules.o RuleSymbol.o \
SyntaxTree.o Word.o NonTerm.o RulePhrase.o AlignedSentenceSyntax.o pugixml.o
extract-mixed-syntax: $(OBJECTS)
g++ $(OBJECTS) -L../../../boost/lib64 -lz -lboost_iostreams-mt -lboost_program_options-mt -o extract-mixed-syntax

65
contrib/other-builds/extract-mixed-syntax/NonTerm.cpp
View File

@ -1,65 +0,0 @@
/*
* NonTerm.cpp
*
* Created on: 22 Feb 2014
* Author: hieu
*/
#include <sstream>
#include "NonTerm.h"
#include "Word.h"
#include "ConsistentPhrase.h"
#include "Parameter.h"
using namespace std;
NonTerm::NonTerm(const ConsistentPhrase &consistentPhrase,
const std::string &source,
const std::string &target)
:m_consistentPhrase(&consistentPhrase)
,m_source(source)
,m_target(target)
{
// TODO Auto-generated constructor stub
}
NonTerm::~NonTerm() {
// TODO Auto-generated destructor stub
}
std::string NonTerm::Debug() const
{
stringstream out;
out << m_source << m_target;
out << m_consistentPhrase->Debug();
return out.str();
}
void NonTerm::Output(std::ostream &out) const
{
out << m_source << m_target;
}
void NonTerm::Output(std::ostream &out, Moses::FactorDirection direction) const
{
out << GetLabel(direction);
}
const std::string &NonTerm::GetLabel(Moses::FactorDirection direction) const
{
return (direction == Moses::Input) ? m_source : m_target;
}
bool NonTerm::IsHiero(Moses::FactorDirection direction, const Parameter &params) const
{
const std::string &label = NonTerm::GetLabel(direction);
return label == params.hieroNonTerm;
}
bool NonTerm::IsHiero(const Parameter &params) const
{
return IsHiero(Moses::Input, params) && IsHiero(Moses::Output, params);
}
int NonTerm::GetWidth(Moses::FactorDirection direction) const
{ return GetConsistentPhrase().GetWidth(direction); }

47
contrib/other-builds/extract-mixed-syntax/NonTerm.h
View File

@ -1,47 +0,0 @@
/*
* NonTerm.h
*
* Created on: 22 Feb 2014
* Author: hieu
*/
#pragma once
#include <string>
#include "RuleSymbol.h"
#include "moses/TypeDef.h"
class ConsistentPhrase;
class Parameter;
class NonTerm : public RuleSymbol
{
public:
NonTerm(const ConsistentPhrase &consistentPhrase,
const std::string &source,
const std::string &target);
virtual ~NonTerm();
const ConsistentPhrase &GetConsistentPhrase() const
{ return *m_consistentPhrase; }
int GetWidth(Moses::FactorDirection direction) const;
virtual bool IsNonTerm() const
{ return true; }
std::string GetString() const
{ return m_source + m_target; }
virtual std::string Debug() const;
virtual void Output(std::ostream &out) const;
void Output(std::ostream &out, Moses::FactorDirection direction) const;
const std::string &GetLabel(Moses::FactorDirection direction) const;
bool IsHiero(Moses::FactorDirection direction, const Parameter &params) const;
bool IsHiero(const Parameter &params) const;
protected:
const ConsistentPhrase *m_consistentPhrase;
std::string m_source, m_target;
};

79
contrib/other-builds/extract-mixed-syntax/OutputFileStream.cpp
View File

@ -1,79 +0,0 @@
// $Id: OutputFileStream.cpp 2780 2010-01-29 17:11:17Z bojar $
/***********************************************************************
Moses - factored phrase-based language decoder
Copyright (C) 2006 University of Edinburgh
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
***********************************************************************/
#include <boost/iostreams/filter/gzip.hpp>
#include "OutputFileStream.h"
#include "gzfilebuf.h"
using namespace std;
namespace Moses
{
OutputFileStream::OutputFileStream()
:boost::iostreams::filtering_ostream()
,m_outFile(NULL)
{
}
OutputFileStream::OutputFileStream(const std::string &filePath)
: m_outFile(NULL)
{
Open(filePath);
}
OutputFileStream::~OutputFileStream()
{
Close();
}
bool OutputFileStream::Open(const std::string &filePath)
{
m_outFile = new ofstream(filePath.c_str(), ios_base::out | ios_base::binary);
if (m_outFile->fail()) {
return false;
}
if (filePath.size() > 3 && filePath.substr(filePath.size() - 3, 3) == ".gz") {
this->push(boost::iostreams::gzip_compressor());
}
this->push(*m_outFile);
return true;
}
void OutputFileStream::Close()
{
if (m_outFile == NULL) {
return;
}
this->flush();
this->pop(); // file
m_outFile->close();
delete m_outFile;
m_outFile = NULL;
return;
}
}

50
contrib/other-builds/extract-mixed-syntax/OutputFileStream.h
View File

@ -1,50 +0,0 @@
// $Id: InputFileStream.h 2939 2010-02-24 11:15:44Z jfouet $
/***********************************************************************
Moses - factored phrase-based language decoder
Copyright (C) 2006 University of Edinburgh
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
***********************************************************************/
#pragma once
#include <cstdlib>
#include <fstream>
#include <string>
#include <iostream>
#include <boost/iostreams/filtering_stream.hpp>
namespace Moses
{
/** Used in place of std::istream, can read zipped files if it ends in .gz
*/
class OutputFileStream : public boost::iostreams::filtering_ostream
{
protected:
std::ofstream *m_outFile;
public:
OutputFileStream();
OutputFileStream(const std::string &filePath);
virtual ~OutputFileStream();
bool Open(const std::string &filePath);
void Close();
};
}

41
contrib/other-builds/extract-mixed-syntax/Parameter.cpp
View File

@ -1,41 +0,0 @@
/*
* Parameter.cpp
*
* Created on: 17 Feb 2014
* Author: hieu
*/
#include "Parameter.h"
Parameter::Parameter()
:maxSpan(10)
,maxNonTerm(2)
,maxHieroNonTerm(999)
,maxSymbolsTarget(999)
,maxSymbolsSource(5)
,minHoleSource(2)
,sentenceOffset(0)
,nonTermConsecSource(false)
,requireAlignedWord(true)
,fractionalCounting(true)
,gzOutput(false)
,hieroNonTerm("[X]")
,sourceSyntax(false)
,targetSyntax(false)
,mixedSyntaxType(0)
,multiLabel(0)
,nonTermConsecSourceMixed(true)
,hieroSourceLHS(false)
,maxSpanFreeNonTermSource(0)
,nieceTerminal(true)
,maxScope(UNDEFINED)
,spanLength(false)
,nonTermContext(false)
{}
Parameter::~Parameter() {
// TODO Auto-generated destructor stub
}

51
contrib/other-builds/extract-mixed-syntax/Parameter.h
View File

@ -1,51 +0,0 @@
/*
* Parameter.h
*
* Created on: 17 Feb 2014
* Author: hieu
*/
#pragma once
#include <string>
#include <limits>
#define UNDEFINED std::numeric_limits<int>::max()
class Parameter
{
public:
Parameter();
virtual ~Parameter();
int maxSpan;
int maxNonTerm;
int maxHieroNonTerm;
int maxSymbolsTarget;
int maxSymbolsSource;
int minHoleSource;
long sentenceOffset;
bool nonTermConsecSource;
bool requireAlignedWord;
bool fractionalCounting;
bool gzOutput;
std::string hieroNonTerm;
std::string gluePath;
bool sourceSyntax, targetSyntax;
int mixedSyntaxType, multiLabel;
bool nonTermConsecSourceMixed;
bool hieroSourceLHS;
int maxSpanFreeNonTermSource;
bool nieceTerminal;
int maxScope;
// prperties
bool spanLength;
bool nonTermContext;
};

14
contrib/other-builds/extract-mixed-syntax/Phrase.cpp
View File

@ -1,14 +0,0 @@
#include <sstream>
#include "Phrase.h"
std::string Phrase::Debug() const
{
std::stringstream out;
for (size_t i = 0; i < size(); ++i) {
Word &word = *at(i);
out << word.Debug() << " ";
}
return out.str();
}

19
contrib/other-builds/extract-mixed-syntax/Phrase.h
View File

@ -1,19 +0,0 @@
#pragma once
#include <vector>
#include "Word.h"
// a vector of terminals
class Phrase : public std::vector<Word*>
{
public:
Phrase()
{}
Phrase(size_t size)
:std::vector<Word*>(size)
{}
std::string Debug() const;
};

540
contrib/other-builds/extract-mixed-syntax/Rule.cpp
View File

@ -1,540 +0,0 @@
/*
* Rule.cpp
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#include <sstream>
#include <algorithm>
#include "Rule.h"
#include "AlignedSentence.h"
#include "ConsistentPhrase.h"
#include "NonTerm.h"
#include "Parameter.h"
using namespace std;
Rule::Rule(const NonTerm &lhsNonTerm, const AlignedSentence &alignedSentence)
:m_lhs(lhsNonTerm)
,m_alignedSentence(alignedSentence)
,m_isValid(true)
,m_canRecurse(true)
{
CreateSource();
}
Rule::Rule(const Rule &copy, const NonTerm &nonTerm)
:m_lhs(copy.m_lhs)
,m_alignedSentence(copy.m_alignedSentence)
,m_isValid(true)
,m_canRecurse(true)
,m_nonterms(copy.m_nonterms)
{
m_nonterms.push_back(&nonTerm);
CreateSource();
}
Rule::~Rule() {
// TODO Auto-generated destructor stub
}
const ConsistentPhrase &Rule::GetConsistentPhrase() const
{ return m_lhs.GetConsistentPhrase(); }
void Rule::CreateSource()
{
const NonTerm *cp = NULL;
size_t nonTermInd = 0;
if (nonTermInd < m_nonterms.size()) {
cp = m_nonterms[nonTermInd];
}
for (int sourcePos = m_lhs.GetConsistentPhrase().corners[0];
sourcePos <= m_lhs.GetConsistentPhrase().corners[1];
++sourcePos) {
const RuleSymbol *ruleSymbol;
if (cp && cp->GetConsistentPhrase().corners[0] <= sourcePos && sourcePos <= cp->GetConsistentPhrase().corners[1]) {
// replace words with non-term
ruleSymbol = cp;
sourcePos = cp->GetConsistentPhrase().corners[1];
if (m_nonterms.size()) {
cp = m_nonterms[nonTermInd];
}
// move to next non-term
++nonTermInd;
cp = (nonTermInd < m_nonterms.size()) ? m_nonterms[nonTermInd] : NULL;
}
else {
// terminal
ruleSymbol = m_alignedSentence.GetPhrase(Moses::Input)[sourcePos];
}
m_source.Add(ruleSymbol);
}
}
int Rule::GetNextSourcePosForNonTerm() const
{
if (m_nonterms.empty()) {
// no non-terms so far. Can start next non-term on left corner
return m_lhs.GetConsistentPhrase().corners[0];
}
else {
// next non-term can start just left of previous
const ConsistentPhrase &cp = m_nonterms.back()->GetConsistentPhrase();
int nextPos = cp.corners[1] + 1;
return nextPos;
}
}
std::string Rule::Debug() const
{
stringstream out;
// source
for (size_t i = 0; i < m_source.GetSize(); ++i) {
const RuleSymbol &symbol = *m_source[i];
out << symbol.Debug() << " ";
}
// target
out << "||| ";
for (size_t i = 0; i < m_target.GetSize(); ++i) {
const RuleSymbol &symbol = *m_target[i];
out << symbol.Debug() << " ";
}
out << "||| ";
Alignments::const_iterator iterAlign;
for (iterAlign = m_alignments.begin(); iterAlign != m_alignments.end(); ++iterAlign) {
const std::pair<int,int> &alignPair = *iterAlign;
out << alignPair.first << "-" << alignPair.second << " ";
}
// overall range
out << "||| LHS=" << m_lhs.Debug();
return out.str();
}
void Rule::Output(std::ostream &out, bool forward, const Parameter &params) const
{
if (forward) {
// source
m_source.Output(out);
m_lhs.Output(out, Moses::Input);
out << " ||| ";
// target
m_target.Output(out);
m_lhs.Output(out, Moses::Output);
}
else {
// target
m_target.Output(out);
m_lhs.Output(out, Moses::Output);
out << " ||| ";
// source
m_source.Output(out);
m_lhs.Output(out, Moses::Input);
}
out << " ||| ";
// alignment
Alignments::const_iterator iterAlign;
for (iterAlign = m_alignments.begin(); iterAlign != m_alignments.end(); ++iterAlign) {
const std::pair<int,int> &alignPair = *iterAlign;
if (forward) {
out << alignPair.first << "-" << alignPair.second << " ";
}
else {
out << alignPair.second << "-" << alignPair.first << " ";
}
}
out << "||| ";
// count
out << m_count;
out << " ||| ";
// properties
// span length
if (forward && params.spanLength && m_nonterms.size()) {
out << "{{SpanLength ";
for (size_t i = 0; i < m_nonterms.size(); ++i) {
const NonTerm &nonTerm = *m_nonterms[i];
const ConsistentPhrase &cp = nonTerm.GetConsistentPhrase();
out << i << "," << cp.GetWidth(Moses::Input) << "," << cp.GetWidth(Moses::Output) << " ";
}
out << "}} ";
}
// non-term context
if (forward && params.nonTermContext && m_nonterms.size()) {
out << "{{NonTermContext ";
for (size_t i = 0; i < m_nonterms.size(); ++i) {
const NonTerm &nonTerm = *m_nonterms[i];
const ConsistentPhrase &cp = nonTerm.GetConsistentPhrase();
NonTermContext(i, cp, out);
}
out << "}} ";
}
}
void Rule::NonTermContext(size_t ntInd, const ConsistentPhrase &cp, std::ostream &out) const
{
int startPos = cp.corners[0];
int endPos = cp.corners[1];
const Phrase &source = m_alignedSentence.GetPhrase(Moses::Input);
if (startPos == 0) {
out << "<s> ";
}
else {
out << source[startPos - 1]->GetString() << " ";
}
out << source[startPos]->GetString() << " ";
out << source[endPos]->GetString() << " ";
if (endPos == source.size() - 1) {
out << "</s> ";
}
else {
out << source[endPos + 1]->GetString() << " ";
}
}
void Rule::Prevalidate(const Parameter &params)
{
const ConsistentPhrase &cp = m_lhs.GetConsistentPhrase();
// check number of source symbols in rule
if (m_source.GetSize() > params.maxSymbolsSource) {
m_isValid = false;
}
// check that last non-term added isn't too small
if (m_nonterms.size()) {
const NonTerm &lastNonTerm = *m_nonterms.back();
const ConsistentPhrase &cp = lastNonTerm.GetConsistentPhrase();
int sourceWidth = cp.corners[1] - cp.corners[0] + 1;
if (sourceWidth < params.minHoleSource) {
m_isValid = false;
m_canRecurse = false;
return;
}
}
// check number of non-terms
int numNonTerms = 0;
int numHieroNonTerms = 0;
for (size_t i = 0; i < m_source.GetSize(); ++i) {
const RuleSymbol *arc = m_source[i];
if (arc->IsNonTerm()) {
++numNonTerms;
const NonTerm &nonTerm = *static_cast<const NonTerm*>(arc);
bool isHiero = nonTerm.IsHiero(params);
if (isHiero) {
++numHieroNonTerms;
}
}
}
if (numNonTerms >= params.maxNonTerm) {
m_canRecurse = false;
if (numNonTerms > params.maxNonTerm) {
m_isValid = false;
return;
}
}
if (numHieroNonTerms >= params.maxHieroNonTerm) {
m_canRecurse = false;
if (numHieroNonTerms > params.maxHieroNonTerm) {
m_isValid = false;
return;
}
}
// check if 2 consecutive non-terms in source
if (!params.nonTermConsecSource && m_nonterms.size() >= 2) {
const NonTerm &lastNonTerm = *m_nonterms.back();
const NonTerm &secondLastNonTerm = *m_nonterms[m_nonterms.size() - 2];
if (secondLastNonTerm.GetConsistentPhrase().corners[1] + 1 ==
lastNonTerm.GetConsistentPhrase().corners[0]) {
if (params.mixedSyntaxType == 0) {
// ordinary hiero or syntax model
m_isValid = false;
m_canRecurse = false;
return;
}
else {
// Hieu's mixed syntax
if (lastNonTerm.IsHiero(Moses::Input, params)
&& secondLastNonTerm.IsHiero(Moses::Input, params)) {
m_isValid = false;
m_canRecurse = false;
return;
}
}
}
}
//check to see if it overlaps with any other non-terms
if (m_nonterms.size() >= 2) {
const NonTerm &lastNonTerm = *m_nonterms.back();
for (size_t i = 0; i < m_nonterms.size() - 1; ++i) {
const NonTerm &otherNonTerm = *m_nonterms[i];
bool overlap = lastNonTerm.GetConsistentPhrase().TargetOverlap(otherNonTerm.GetConsistentPhrase());
if (overlap) {
m_isValid = false;
m_canRecurse = false;
return;
}
}
}
// check that at least 1 word is aligned
if (params.requireAlignedWord) {
bool ok = false;
for (size_t i = 0; i < m_source.GetSize(); ++i) {
const RuleSymbol &symbol = *m_source[i];
if (!symbol.IsNonTerm()) {
const Word &word = static_cast<const Word&>(symbol);
if (word.GetAlignment().size()) {
ok = true;
break;
}
}
}
if (!ok) {
m_isValid = false;
m_canRecurse = false;
return;
}
}
if (params.maxSpanFreeNonTermSource) {
const NonTerm *front = dynamic_cast<const NonTerm*>(m_source[0]);
if (front) {
int width = front->GetWidth(Moses::Input);
if (width > params.maxSpanFreeNonTermSource) {
m_isValid = false;
m_canRecurse = false;
return;
}
}
const NonTerm *back = dynamic_cast<const NonTerm*>(m_source.Back());
if (back) {
int width = back->GetWidth(Moses::Input);
if (width > params.maxSpanFreeNonTermSource) {
m_isValid = false;
m_canRecurse = false;
return;
}
}
}
if (!params.nieceTerminal) {
// collect terminal in a rule
std::set<const Word*> terms;
for (size_t i = 0; i < m_source.GetSize(); ++i) {
const Word *word = dynamic_cast<const Word*>(m_source[i]);
if (word) {
terms.insert(word);
}
}
// look in non-terms
for (size_t i = 0; i < m_source.GetSize(); ++i) {
const NonTerm *nonTerm = dynamic_cast<const NonTerm*>(m_source[i]);
if (nonTerm) {
const ConsistentPhrase &cp = nonTerm->GetConsistentPhrase();
bool containTerm = ContainTerm(cp, terms);
if (containTerm) {
//cerr << "ruleSource=" << *ruleSource << " ";
//cerr << "ntRange=" << ntRange << endl;
// non-term contains 1 of the terms in the rule.
m_isValid = false;
m_canRecurse = false;
return;
}
}
}
}
if (params.maxScope != UNDEFINED) {
int scope = CalcScope();
if (scope > params.maxScope) {
m_isValid = false;
m_canRecurse = false;
return;
}
}
}
int Rule::CalcScope() const
{
int scope = 0;
if (m_source.GetSize() > 1) {
const RuleSymbol &front = *m_source.Front();
if (front.IsNonTerm()) {
++scope;
}
const RuleSymbol &back = *m_source.Back();
if (back.IsNonTerm()) {
++scope;
}
}
return scope;
}
template<typename T>
bool Contains(const T *sought, const set<const T*> &coll)
{
std::set<const Word*>::const_iterator iter;
for (iter = coll.begin(); iter != coll.end(); ++iter) {
const Word *found = *iter;
if (sought->CompareString(*found) == 0) {
return true;
}
}
return false;
}
bool Rule::ContainTerm(const ConsistentPhrase &cp, const std::set<const Word*> &terms) const
{
const Phrase &sourceSentence = m_alignedSentence.GetPhrase(Moses::Input);
for (int pos = cp.corners[0]; pos <= cp.corners[1]; ++pos) {
const Word *soughtWord = sourceSentence[pos];
// find same word in set
if (Contains(soughtWord, terms)) {
return true;
}
}
return false;
}
bool CompareTargetNonTerms(const NonTerm *a, const NonTerm *b)
{
// compare just start target pos
return a->GetConsistentPhrase().corners[2] < b->GetConsistentPhrase().corners[2];
}
void Rule::CreateTarget(const Parameter &params)
{
if (!m_isValid) {
return;
}
vector<const NonTerm*> targetNonTerm(m_nonterms);
std::sort(targetNonTerm.begin(), targetNonTerm.end(), CompareTargetNonTerms);
const NonTerm *cp = NULL;
size_t nonTermInd = 0;
if (nonTermInd < targetNonTerm.size()) {
cp = targetNonTerm[nonTermInd];
}
for (int targetPos = m_lhs.GetConsistentPhrase().corners[2];
targetPos <= m_lhs.GetConsistentPhrase().corners[3];
++targetPos) {
const RuleSymbol *ruleSymbol;
if (cp && cp->GetConsistentPhrase().corners[2] <= targetPos && targetPos <= cp->GetConsistentPhrase().corners[3]) {
// replace words with non-term
ruleSymbol = cp;
targetPos = cp->GetConsistentPhrase().corners[3];
if (targetNonTerm.size()) {
cp = targetNonTerm[nonTermInd];
}
// move to next non-term
++nonTermInd;
cp = (nonTermInd < targetNonTerm.size()) ? targetNonTerm[nonTermInd] : NULL;
}
else {
// terminal
ruleSymbol = m_alignedSentence.GetPhrase(Moses::Output)[targetPos];
}
m_target.Add(ruleSymbol);
}
CreateAlignments();
}
void Rule::CreateAlignments()
{
int sourceStart = GetConsistentPhrase().corners[0];
int targetStart = GetConsistentPhrase().corners[2];
for (size_t sourcePos = 0; sourcePos < m_source.GetSize(); ++sourcePos) {
const RuleSymbol *symbol = m_source[sourcePos];
if (!symbol->IsNonTerm()) {
// terminals
const Word &sourceWord = static_cast<const Word&>(*symbol);
const std::set<const Word *> &targetWords = sourceWord.GetAlignment();
CreateAlignments(sourcePos, targetWords);
}
else {
// non-terms. same object in both source & target
CreateAlignments(sourcePos, symbol);
}
}
}
void Rule::CreateAlignments(int sourcePos, const std::set<const Word *> &targetWords)
{
std::set<const Word *>::const_iterator iterTarget;
for (iterTarget = targetWords.begin(); iterTarget != targetWords.end(); ++iterTarget) {
const Word *targetWord = *iterTarget;
CreateAlignments(sourcePos, targetWord);
}
}
void Rule::CreateAlignments(int sourcePos, const RuleSymbol *targetSought)
{
// should be in target phrase
for (size_t targetPos = 0; targetPos < m_target.GetSize(); ++targetPos) {
const RuleSymbol *foundSymbol = m_target[targetPos];
if (targetSought == foundSymbol) {
pair<int, int> alignPoint(sourcePos, targetPos);
m_alignments.insert(alignPoint);
return;
}
}
throw "not found";
}

87
contrib/other-builds/extract-mixed-syntax/Rule.h
View File

@ -1,87 +0,0 @@
/*
* Rule.h
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#pragma once
#include <vector>
#include "Phrase.h"
#include "RulePhrase.h"
#include "moses/TypeDef.h"
class ConsistentPhrase;
class AlignedSentence;
class NonTerm;
class Parameter;
class Rule {
public:
typedef std::set<std::pair<int,int> > Alignments;
Rule(const Rule &copy); // do not implement
// original rule with no non-term
Rule(const NonTerm &lhsNonTerm, const AlignedSentence &alignedSentence);
// extend a rule, adding 1 new non-term
Rule(const Rule &copy, const NonTerm &nonTerm);
virtual ~Rule();
bool IsValid() const
{ return m_isValid; }
bool CanRecurse() const
{ return m_canRecurse; }
const NonTerm &GetLHS() const
{ return m_lhs; }
const ConsistentPhrase &GetConsistentPhrase() const;
int GetNextSourcePosForNonTerm() const;
void SetCount(float count)
{ m_count = count; }
float GetCount() const
{ return m_count; }
const Alignments &GetAlignments() const
{ return m_alignments; }
std::string Debug() const;
void Output(std::ostream &out, bool forward, const Parameter &params) const;
void Prevalidate(const Parameter &params);
void CreateTarget(const Parameter &params);
const RulePhrase &GetPhrase(Moses::FactorDirection direction) const
{ return (direction == Moses::Input) ? m_source : m_target; }
protected:
const NonTerm &m_lhs;
const AlignedSentence &m_alignedSentence;
RulePhrase m_source, m_target;
float m_count;
Alignments m_alignments;
// in source order
std::vector<const NonTerm*> m_nonterms;
bool m_isValid, m_canRecurse;
void CreateSource();
void CreateAlignments();
void CreateAlignments(int sourcePos, const std::set<const Word *> &targetWords);
void CreateAlignments(int sourcePos, const RuleSymbol *targetSought);
bool ContainTerm(const ConsistentPhrase &cp, const std::set<const Word*> &terms) const;
int CalcScope() const; // not yet correctly calculated
void NonTermContext(size_t ntInd, const ConsistentPhrase &cp, std::ostream &out) const;
};

50
contrib/other-builds/extract-mixed-syntax/RulePhrase.cpp
View File

@ -1,50 +0,0 @@
/*
* RulePhrase.cpp
*
* Created on: 26 Feb 2014
* Author: hieu
*/
#include <sstream>
#include "RulePhrase.h"
#include "RuleSymbol.h"
using namespace std;
extern bool g_debug;
int RulePhrase::Compare(const RulePhrase &other) const
{
if (GetSize() != other.GetSize()) {
return GetSize() < other.GetSize() ? -1 : +1;
}
for (size_t i = 0; i < m_coll.size(); ++i) {
const RuleSymbol &symbol = *m_coll[i];
const RuleSymbol &otherSymbol = *other.m_coll[i];
int compare = symbol.Compare(otherSymbol);
if (compare) {
return compare;
}
}
return 0;
}
void RulePhrase::Output(std::ostream &out) const
{
for (size_t i = 0; i < m_coll.size(); ++i) {
const RuleSymbol &symbol = *m_coll[i];
symbol.Output(out);
out << " ";
}
}
std::string RulePhrase::Debug() const
{
std::stringstream out;
Output(out);
return out.str();
}

49
contrib/other-builds/extract-mixed-syntax/RulePhrase.h
View File

@ -1,49 +0,0 @@
/*
* RulePhrase.h
*
* Created on: 26 Feb 2014
* Author: hieu
*/
#ifndef RULEPHRASE_H_
#define RULEPHRASE_H_
#include <vector>
#include <cstddef>
#include <iostream>
class RuleSymbol;
// a phrase of terms and non-terms for 1 side of a rule
class RulePhrase
{
public:
typedef std::vector<const RuleSymbol*> Coll;
Coll m_coll;
size_t GetSize() const
{ return m_coll.size(); }
void Add(const RuleSymbol *symbol)
{
m_coll.push_back(symbol);
}
const RuleSymbol* operator[](size_t index) const {
return m_coll[index];
}
const RuleSymbol* Front() const {
return m_coll.front();
}
const RuleSymbol* Back() const {
return m_coll.back();
}
int Compare(const RulePhrase &other) const;
void Output(std::ostream &out) const;
std::string Debug() const;
};
#endif /* RULEPHRASE_H_ */

36
contrib/other-builds/extract-mixed-syntax/RuleSymbol.cpp
View File

@ -1,36 +0,0 @@
/*
* RuleSymbol.cpp
*
* Created on: 21 Feb 2014
* Author: hieu
*/
#include "RuleSymbol.h"
using namespace std;
RuleSymbol::RuleSymbol() {
// TODO Auto-generated constructor stub
}
RuleSymbol::~RuleSymbol() {
// TODO Auto-generated destructor stub
}
int RuleSymbol::Compare(const RuleSymbol &other) const
{
if (IsNonTerm() != other.IsNonTerm()) {
return IsNonTerm() ? -1 : +1;
}
string str = GetString();
string otherStr = other.GetString();
if (str == otherStr) {
return 0;
}
else {
return (str < otherStr) ? -1 : +1;
}
}

31
contrib/other-builds/extract-mixed-syntax/RuleSymbol.h
View File

@ -1,31 +0,0 @@
/*
* RuleSymbol.h
*
* Created on: 21 Feb 2014
* Author: hieu
*/
#ifndef RULESYMBOL_H_
#define RULESYMBOL_H_
#include <iostream>
#include <string>
// base class - terminal or non-term
class RuleSymbol {
public:
RuleSymbol();
virtual ~RuleSymbol();
virtual bool IsNonTerm() const = 0;
virtual std::string Debug() const = 0;
virtual void Output(std::ostream &out) const = 0;
virtual std::string GetString() const = 0;
int Compare(const RuleSymbol &other) const;
};
#endif /* RULESYMBOL_H_ */

227
contrib/other-builds/extract-mixed-syntax/Rules.cpp
View File

@ -1,227 +0,0 @@
/*
* Rules.cpp
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#include <sstream>
#include "Rules.h"
#include "ConsistentPhrase.h"
#include "ConsistentPhrases.h"
#include "AlignedSentence.h"
#include "Rule.h"
#include "Parameter.h"
#include "moses/Util.h"
using namespace std;
extern bool g_debug;
Rules::Rules(const AlignedSentence &alignedSentence)
:m_alignedSentence(alignedSentence)
{
}
Rules::~Rules() {
Moses::RemoveAllInColl(m_keepRules);
}
void Rules::CreateRules(const ConsistentPhrase &cp,
const Parameter &params)
{
if (params.hieroSourceLHS) {
const NonTerm &nonTerm = cp.GetHieroNonTerm();
CreateRule(nonTerm, params);
}
else {
const ConsistentPhrase::NonTerms &nonTerms = cp.GetNonTerms();
for (size_t i = 0; i < nonTerms.size(); ++i) {
const NonTerm &nonTerm = nonTerms[i];
CreateRule(nonTerm, params);
}
}
}
void Rules::CreateRule(const NonTerm &nonTerm,
const Parameter &params)
{
Rule *rule = new Rule(nonTerm, m_alignedSentence);
rule->Prevalidate(params);
rule->CreateTarget(params);
if (rule->CanRecurse()) {
Extend(*rule, params);
}
if (rule->IsValid()) {
m_keepRules.insert(rule);
}
else {
delete rule;
}
}
void Rules::Extend(const Parameter &params)
{
const ConsistentPhrases &allCPS = m_alignedSentence.GetConsistentPhrases();
size_t size = m_alignedSentence.GetPhrase(Moses::Input).size();
for (size_t sourceStart = 0; sourceStart < size; ++sourceStart) {
for (size_t sourceEnd = sourceStart; sourceEnd < size; ++sourceEnd) {
const ConsistentPhrases::Coll &cps = allCPS.GetColl(sourceStart, sourceEnd);
ConsistentPhrases::Coll::const_iterator iter;
for (iter = cps.begin(); iter != cps.end(); ++iter) {
const ConsistentPhrase &cp = **iter;
CreateRules(cp, params);
}
}
}
}
void Rules::Extend(const Rule &rule, const Parameter &params)
{
const ConsistentPhrases &allCPS = m_alignedSentence.GetConsistentPhrases();
int sourceMin = rule.GetNextSourcePosForNonTerm();
int ruleStart = rule.GetConsistentPhrase().corners[0];
int ruleEnd = rule.GetConsistentPhrase().corners[1];
for (int sourceStart = sourceMin; sourceStart <= ruleEnd; ++sourceStart) {
for (int sourceEnd = sourceStart; sourceEnd <= ruleEnd; ++sourceEnd) {
if (sourceStart == ruleStart && sourceEnd == ruleEnd) {
// don't cover whole rule with 1 non-term
continue;
}
const ConsistentPhrases::Coll &cps = allCPS.GetColl(sourceStart, sourceEnd);
Extend(rule, cps, params);
}
}
}
void Rules::Extend(const Rule &rule, const ConsistentPhrases::Coll &cps, const Parameter &params)
{
ConsistentPhrases::Coll::const_iterator iter;
for (iter = cps.begin(); iter != cps.end(); ++iter) {
const ConsistentPhrase &cp = **iter;
Extend(rule, cp, params);
}
}
void Rules::Extend(const Rule &rule, const ConsistentPhrase &cp, const Parameter &params)
{
const ConsistentPhrase::NonTerms &nonTerms = cp.GetNonTerms();
for (size_t i = 0; i < nonTerms.size(); ++i) {
const NonTerm &nonTerm = nonTerms[i];
Rule *newRule = new Rule(rule, nonTerm);
newRule->Prevalidate(params);
newRule->CreateTarget(params);
if (newRule->CanRecurse()) {
// recursively extend
Extend(*newRule, params);
}
if (newRule->IsValid()) {
m_keepRules.insert(newRule);
}
else {
delete newRule;
}
}
}
std::string Rules::Debug() const
{
stringstream out;
std::set<Rule*>::const_iterator iter;
out << "m_keepRules:" << endl;
for (iter = m_keepRules.begin(); iter != m_keepRules.end(); ++iter) {
const Rule &rule = **iter;
out << rule.Debug() << endl;
}
return out.str();
}
void Rules::Output(std::ostream &out, bool forward, const Parameter &params) const
{
std::set<Rule*, CompareRules>::const_iterator iter;
for (iter = m_mergeRules.begin(); iter != m_mergeRules.end(); ++iter) {
const Rule &rule = **iter;
rule.Output(out, forward, params);
out << endl;
}
}
void Rules::Consolidate(const Parameter &params)
{
if (params.fractionalCounting) {
CalcFractionalCount();
}
else {
std::set<Rule*>::iterator iter;
for (iter = m_keepRules.begin(); iter != m_keepRules.end(); ++iter) {
Rule &rule = **iter;
rule.SetCount(1);
}
}
MergeRules(params);
}
void Rules::MergeRules(const Parameter &params)
{
typedef std::set<Rule*, CompareRules> MergeRules;
std::set<Rule*>::const_iterator iterOrig;
for (iterOrig = m_keepRules.begin(); iterOrig != m_keepRules.end(); ++iterOrig) {
Rule *origRule = *iterOrig;
pair<MergeRules::iterator, bool> inserted = m_mergeRules.insert(origRule);
if (!inserted.second) {
// already there, just add count
Rule &rule = **inserted.first;
float newCount = rule.GetCount() + origRule->GetCount();
rule.SetCount(newCount);
}
}
}
void Rules::CalcFractionalCount()
{
typedef std::set<Rule*> RuleColl;
typedef std::map<const ConsistentPhrase*, RuleColl> RuleByConsistentPhrase;
RuleByConsistentPhrase allRules;
// sort by source AND target ranges
std::set<Rule*>::const_iterator iter;
for (iter = m_keepRules.begin(); iter != m_keepRules.end(); ++iter) {
Rule *rule = *iter;
const ConsistentPhrase &cp = rule->GetConsistentPhrase();
RuleColl &ruleColl = allRules[&cp];
ruleColl.insert(rule);
}
// fractional count
RuleByConsistentPhrase::iterator iterOuter;
for (iterOuter = allRules.begin(); iterOuter != allRules.end(); ++iterOuter) {
RuleColl &rules = iterOuter->second;
RuleColl::iterator iterInner;
for (iterInner = rules.begin(); iterInner != rules.end(); ++iterInner) {
Rule &rule = **iterInner;
rule.SetCount(1.0f / (float) rules.size());
}
}
}

72
contrib/other-builds/extract-mixed-syntax/Rules.h
View File

@ -1,72 +0,0 @@
/*
* Rules.h
*
* Created on: 20 Feb 2014
* Author: hieu
*/
#pragma once
#include <set>
#include <iostream>
#include "ConsistentPhrases.h"
#include "Rule.h"
extern bool g_debug;
class AlignedSentence;
class Parameter;
struct CompareRules {
bool operator()(const Rule *a, const Rule *b)
{
int compare;
compare = a->GetPhrase(Moses::Input).Compare(b->GetPhrase(Moses::Input));
if (compare) return compare < 0;
compare = a->GetPhrase(Moses::Output).Compare(b->GetPhrase(Moses::Output));
if (compare) return compare < 0;
if (a->GetAlignments() != b->GetAlignments()) {
return a->GetAlignments() < b->GetAlignments();
}
if (a->GetLHS().GetString() != b->GetLHS().GetString()) {
return a->GetLHS().GetString() < b->GetLHS().GetString();
}
return false;
}
};
class Rules {
public:
Rules(const AlignedSentence &alignedSentence);
virtual ~Rules();
void Extend(const Parameter &params);
void Consolidate(const Parameter &params);
std::string Debug() const;
void Output(std::ostream &out, bool forward, const Parameter &params) const;
protected:
const AlignedSentence &m_alignedSentence;
std::set<Rule*> m_keepRules;
std::set<Rule*, CompareRules> m_mergeRules;
void Extend(const Rule &rule, const Parameter &params);
void Extend(const Rule &rule, const ConsistentPhrases::Coll &cps, const Parameter &params);
void Extend(const Rule &rule, const ConsistentPhrase &cp, const Parameter &params);
// create original rules
void CreateRules(const ConsistentPhrase &cp,
const Parameter &params);
void CreateRule(const NonTerm &nonTerm,
const Parameter &params);
void MergeRules(const Parameter &params);
void CalcFractionalCount();
};

47
contrib/other-builds/extract-mixed-syntax/SyntaxTree.cpp
View File

@ -1,47 +0,0 @@
#include <cassert>
#include <iostream>
#include "SyntaxTree.h"
#include "Parameter.h"
using namespace std;
void SyntaxTree::Add(int startPos, int endPos, const std::string &label, const Parameter &params)
{
//cerr << "add " << label << " to " << "[" << startPos << "-" << endPos << "]" << endl;
Range range(startPos, endPos);
Labels &labels = m_coll[range];
bool add = true;
if (labels.size()) {
if (params.multiLabel == 1) {
// delete the label in collection and add new
assert(labels.size() == 1);
labels.clear();
}
else if (params.multiLabel == 2) {
// ignore this label
add = false;
}
}
if (add) {
labels.push_back(label);
}
}
void SyntaxTree::AddToAll(const std::string &label)
{
Coll::iterator iter;
for (iter = m_coll.begin(); iter != m_coll.end(); ++iter) {
Labels &labels = iter->second;
labels.push_back(label);
}
}
const SyntaxTree::Labels &SyntaxTree::Find(int startPos, int endPos) const
{
Coll::const_iterator iter;
iter = m_coll.find(Range(startPos, endPos));
return (iter == m_coll.end()) ? m_defaultLabels : iter->second;
}

32
contrib/other-builds/extract-mixed-syntax/SyntaxTree.h
View File

@ -1,32 +0,0 @@
#pragma once
#include <vector>
#include <map>
#include <string>
class Parameter;
class SyntaxTree
{
public:
typedef std::pair<int, int> Range;
typedef std::vector<std::string> Labels;
typedef std::map<Range, Labels> Coll;
void Add(int startPos, int endPos, const std::string &label, const Parameter &params);
void AddToAll(const std::string &label);
const Labels &Find(int startPos, int endPos) const;
void SetHieroLabel(const std::string &label) {
m_defaultLabels.push_back(label);
}
protected:
Coll m_coll;
Labels m_defaultLabels;
};

56
contrib/other-builds/extract-mixed-syntax/Word.cpp
View File

@ -1,56 +0,0 @@
/*
* Word.cpp
*
* Created on: 18 Feb 2014
* Author: s0565741
*/
#include <limits>
#include "Word.h"
using namespace std;
Word::Word(int pos, const std::string &str)
:m_pos(pos)
,m_str(str)
{
// TODO Auto-generated constructor stub
}
Word::~Word() {
// TODO Auto-generated destructor stub
}
void Word::AddAlignment(const Word *other)
{
m_alignment.insert(other);
}
std::set<int> Word::GetAlignmentIndex() const
{
std::set<int> ret;
std::set<const Word *>::const_iterator iter;
for (iter = m_alignment.begin(); iter != m_alignment.end(); ++iter) {
const Word &otherWord = **iter;
int otherPos = otherWord.GetPos();
ret.insert(otherPos);
}
return ret;
}
void Word::Output(std::ostream &out) const
{
out << m_str;
}
std::string Word::Debug() const
{
return m_str;
}
int Word::CompareString(const Word &other) const
{
return m_str.compare(other.m_str);
}

47
contrib/other-builds/extract-mixed-syntax/Word.h
View File

@ -1,47 +0,0 @@
/*
* Word.h
*
* Created on: 18 Feb 2014
* Author: s0565741
*/
#pragma once
#include <string>
#include <set>
#include "RuleSymbol.h"
// a terminal
class Word : public RuleSymbol
{
public:
Word(const Word&); // do not implement
Word(int pos, const std::string &str);
virtual ~Word();
virtual bool IsNonTerm() const
{ return false; }
std::string GetString() const
{ return m_str; }
int GetPos() const
{ return m_pos; }
void AddAlignment(const Word *other);
const std::set<const Word *> &GetAlignment() const
{ return m_alignment; }
std::set<int> GetAlignmentIndex() const;
void Output(std::ostream &out) const;
std::string Debug() const;
int CompareString(const Word &other) const;
protected:
int m_pos; // original position in sentence, NOT in lattice
std::string m_str;
std::set<const Word *> m_alignment;
};

27
contrib/other-builds/extract-mixed-syntax/filter-by-source-word-count.perl
View File

@ -1,27 +0,0 @@
#!/usr/bin/perl
use strict;
binmode(STDIN, ":utf8");
binmode(STDOUT, ":utf8");
binmode(STDERR, ":utf8");
my $maxNumWords = $ARGV[0];
while (my $line = <STDIN>) {
chomp($line);
my @toks = split(/ /,$line);
my $numSourceWords = 0;
my $tok = $toks[$numSourceWords];
while ($tok ne "|||") {
++$numSourceWords;
$tok = $toks[$numSourceWords];
}
if ($numSourceWords <= $maxNumWords) {
print "$line\n";
}
}

81
contrib/other-builds/extract-mixed-syntax/gzfilebuf.h
View File

@ -1,81 +0,0 @@
#ifndef moses_gzfile_buf_h
#define moses_gzfile_buf_h
#include <streambuf>
#include <zlib.h>
#include <cstring>
class gzfilebuf : public std::streambuf {
public:
gzfilebuf(const char *filename)
{ _gzf = gzopen(filename, "rb");
setg (_buff+sizeof(int), // beginning of putback area
_buff+sizeof(int), // read position
_buff+sizeof(int)); // end position
}
~gzfilebuf() { gzclose(_gzf); }
protected:
virtual int_type overflow (int_type c) {
throw;
}
// write multiple characters
virtual
std::streamsize xsputn (const char* s,
std::streamsize num) {
throw;
}
virtual std::streampos seekpos ( std::streampos sp, std::ios_base::openmode which = std::ios_base::in | std::ios_base::out ){ throw;
}
//read one character
virtual int_type underflow () {
// is read position before end of _buff?
if (gptr() < egptr()) {
return traits_type::to_int_type(*gptr());
}
/* process size of putback area
* - use number of characters read
* - but at most four
*/
unsigned int numPutback = gptr() - eback();
if (numPutback > sizeof(int)) {
numPutback = sizeof(int);
}
/* copy up to four characters previously read into
* the putback _buff (area of first four characters)
*/
std::memmove (_buff+(sizeof(int)-numPutback), gptr()-numPutback,
numPutback);
// read new characters
int num = gzread(_gzf, _buff+sizeof(int), _buffsize-sizeof(int));
if (num <= 0) {
// ERROR or EOF
return EOF;
}
// reset _buff pointers
setg (_buff+(sizeof(int)-numPutback), // beginning of putback area
_buff+sizeof(int), // read position
_buff+sizeof(int)+num); // end of buffer
// return next character
return traits_type::to_int_type(*gptr());
}
std::streamsize xsgetn (char* s,
std::streamsize num) {
return gzread(_gzf,s,num);
}
private:
gzFile _gzf;
static const unsigned int _buffsize = 1024;
char _buff[_buffsize];
};
#endif

33
contrib/other-builds/extract-mixed-syntax/learnable/equal.perl
View File

@ -1,33 +0,0 @@
#! /usr/bin/perl -w
use strict;
sub trim($);
my $file1 = $ARGV[0];
my $file2 = $ARGV[1];
open (FILE1, $file1);
open (FILE2, $file2);
my $countEqual = 0;
while (my $line1 = <FILE1>) {
my $line2 = <FILE2>;
if (trim($line1) eq trim($line2)) {
++$countEqual;
}
}
print $countEqual ."\n";
######################
# Perl trim function to remove whitespace from the start and end of the string
sub trim($) {
my $string = shift;
$string =~ s/^\s+//;
$string =~ s/\s+$//;
return $string;
}

29
contrib/other-builds/extract-mixed-syntax/learnable/get-by-line-number.perl
View File

@ -1,29 +0,0 @@
#! /usr/bin/perl -w
use strict;
binmode(STDIN, ":utf8");
binmode(STDOUT, ":utf8");
binmode(STDERR, ":utf8");
my $fileLineNum = $ARGV[0];
open (FILE_LINE_NUM, $fileLineNum);
my $nextLineNum = <FILE_LINE_NUM>;
my $lineNum = 1;
while (my $line = <STDIN>) {
if (defined($nextLineNum) && $lineNum == $nextLineNum) {
# matches. output line
chomp($line);
print "$line\n";
# next line number
$nextLineNum = <FILE_LINE_NUM>;
}
++$lineNum;
}

108
contrib/other-builds/extract-mixed-syntax/learnable/learnable.perl
View File

@ -1,108 +0,0 @@
#! /usr/bin/perl -w
use strict;
my $iniPath = $ARGV[0];
my $isHiero = $ARGV[1];
my $decoderExec = $ARGV[2];
my $extractExec = $ARGV[3];
my $tmpName = $ARGV[4];
my $WORK_DIR = `pwd`;
chomp($WORK_DIR);
my $MOSES_DIR = "~/workspace/github/mosesdecoder.hieu";
$decoderExec = "$MOSES_DIR/bin/$decoderExec";
$extractExec = "$MOSES_DIR/bin/$extractExec";
my $SPLIT_EXEC = `gsplit --help 2>/dev/null`;
if($SPLIT_EXEC) {
$SPLIT_EXEC = 'gsplit';
}
else {
$SPLIT_EXEC = 'split';
}
my $SORT_EXEC = `gsort --help 2>/dev/null`;
if($SORT_EXEC) {
$SORT_EXEC = 'gsort';
}
else {
$SORT_EXEC = 'sort';
}
my $hieroFlag = "";
if ($isHiero == 1) {
$hieroFlag = "--Hierarchical";
}
print STDERR "WORK_DIR=$WORK_DIR \n";
my $cmd;
open (SOURCE, "source");
open (TARGET, "target");
open (ALIGNMENT, "alignment");
my $lineNum = 0;
my ($source, $target, $alignment);
while ($source = <SOURCE>) {
chomp($source);
$target = <TARGET>; chomp($target);
$alignment = <ALIGNMENT>; chomp($alignment);
#print STDERR "$source ||| $target ||| $alignment \n";
# write out 1 line
my $tmpDir = "$WORK_DIR/$tmpName/work$lineNum";
`mkdir -p $tmpDir`;
open (SOURCE1, ">$tmpDir/source");
open (TARGET1, ">$tmpDir/target");
open (ALIGNMENT1, ">$tmpDir/alignment");
print SOURCE1 "$source\n";
print TARGET1 "$target\n";
print ALIGNMENT1 "$alignment\n";
close (SOURCE1);
close (TARGET1);
close (ALIGNMENT1);
# train
if ($isHiero == 1) {
$cmd = "$extractExec $tmpDir/target $tmpDir/source $tmpDir/alignment $tmpDir/extract --GZOutput";
}
else {
# pb
$cmd = "$extractExec $tmpDir/target $tmpDir/source $tmpDir/alignment $tmpDir/extract 7 --GZOutput";
}
$cmd = "$MOSES_DIR/scripts/generic/extract-parallel.perl 1 $SPLIT_EXEC $SORT_EXEC $cmd";
print STDERR "Executing: $cmd\n";
`$cmd`;
$cmd = "$MOSES_DIR/scripts/generic/score-parallel.perl 1 $SORT_EXEC $MOSES_DIR/bin/score $tmpDir/extract.sorted.gz /dev/null $tmpDir/pt.half.gz $hieroFlag --NoLex 1";
`$cmd`;
$cmd = "$MOSES_DIR/scripts/generic/score-parallel.perl 1 $SORT_EXEC $MOSES_DIR/bin/score $tmpDir/extract.inv.sorted.gz /dev/null $tmpDir/pt.half.inv.gz --Inverse $hieroFlag --NoLex 1";
`$cmd`;
$cmd = "$MOSES_DIR/bin/consolidate $tmpDir/pt.half.gz $tmpDir/pt.half.inv.gz $tmpDir/pt $hieroFlag --OnlyDirect";
`$cmd`;
# decode
$cmd = "$decoderExec -f $iniPath -feature-overwrite \"TranslationModel0 path=$tmpDir/pt\" -i $tmpDir/source -feature-add \"ConstrainedDecoding path=$tmpDir/target\"";
print STDERR "Executing: $cmd\n";
`$cmd`;
# `rm -rf $tmpDir`;
++$lineNum;
}
close(SOURCE);
close(TARGET);
close(ALIGNMENT);

151
contrib/other-builds/extract-mixed-syntax/learnable/num-deriv.perl
View File

@ -1,151 +0,0 @@
#! /usr/bin/perl -w
use strict;
sub Write1Line;
sub WriteCorpus1Holdout;
my $iniPath = $ARGV[0];
my $isHiero = $ARGV[1];
my $decoderExec = $ARGV[2];
my $extractExec = $ARGV[3];
my $tmpName = $ARGV[4];
my $startLine = $ARGV[5];
my $endLine = $ARGV[6];
print STDERR "iniPath=$iniPath \n isHiero=$isHiero \n decoderExec=$decoderExec \n extractExec=$extractExec \n";
my $WORK_DIR = `pwd`;
chomp($WORK_DIR);
my $MOSES_DIR = "~/workspace/github/mosesdecoder.hieu.gna";
$decoderExec = "$MOSES_DIR/bin/$decoderExec";
$extractExec = "$MOSES_DIR/bin/$extractExec";
my $SPLIT_EXEC = `gsplit --help 2>/dev/null`;
if($SPLIT_EXEC) {
$SPLIT_EXEC = 'gsplit';
}
else {
$SPLIT_EXEC = 'split';
}
my $SORT_EXEC = `gsort --help 2>/dev/null`;
if($SORT_EXEC) {
$SORT_EXEC = 'gsort';
}
else {
$SORT_EXEC = 'sort';
}
my $hieroFlag = "";
if ($isHiero == 1) {
$hieroFlag = "--Hierarchical";
}
print STDERR "WORK_DIR=$WORK_DIR \n";
my $cmd;
open (SOURCE, "source");
open (TARGET, "target");
open (ALIGNMENT, "alignment");
my $numLines = `cat source | wc -l`;
for (my $lineNum = 0; $lineNum < $numLines; ++$lineNum) {
my $source = <SOURCE>; chomp($source);
my $target = <TARGET>; chomp($target);
my $alignment = <ALIGNMENT>; chomp($alignment);
if ($lineNum < $startLine || $lineNum >= $endLine) {
next;
}
#print STDERR "$source ||| $target ||| $alignment \n";
# write out 1 line
my $tmpDir = "$WORK_DIR/$tmpName/work$lineNum";
`mkdir -p $tmpDir`;
Write1Line($source, $tmpDir, "source.1");
Write1Line($target, $tmpDir, "target.1");
Write1Line($alignment, $tmpDir, "alignment.1");
WriteCorpus1Holdout($lineNum, "source", $tmpDir, "source.corpus");
WriteCorpus1Holdout($lineNum, "target", $tmpDir, "target.corpus");
WriteCorpus1Holdout($lineNum, "alignment", $tmpDir, "alignment.corpus");
# train
if ($isHiero == 1) {
$cmd = "$extractExec $tmpDir/target.corpus $tmpDir/source.corpus $tmpDir/alignment.corpus $tmpDir/extract --GZOutput";
}
else {
# pb
$cmd = "$extractExec $tmpDir/target.corpus $tmpDir/source.corpus $tmpDir/alignment.corpus $tmpDir/extract 7 --GZOutput";
}
$cmd = "$MOSES_DIR/scripts/generic/extract-parallel.perl 1 $SPLIT_EXEC $SORT_EXEC $cmd";
print STDERR "Executing: $cmd\n";
`$cmd`;
$cmd = "$MOSES_DIR/scripts/generic/score-parallel.perl 1 $SORT_EXEC $MOSES_DIR/bin/score $tmpDir/extract.sorted.gz /dev/null $tmpDir/pt.half.gz $hieroFlag --NoLex 1";
`$cmd`;
$cmd = "$MOSES_DIR/scripts/generic/score-parallel.perl 1 $SORT_EXEC $MOSES_DIR/bin/score $tmpDir/extract.inv.sorted.gz /dev/null $tmpDir/pt.half.inv.gz --Inverse $hieroFlag --NoLex 1";
`$cmd`;
$cmd = "$MOSES_DIR/bin/consolidate $tmpDir/pt.half.gz $tmpDir/pt.half.inv.gz $tmpDir/pt $hieroFlag --OnlyDirect";
`$cmd`;
# decode
$cmd = "$decoderExec -f $iniPath -feature-overwrite \"TranslationModel0 path=$tmpDir/pt\" -i $tmpDir/source.1 -n-best-list $tmpDir/nbest 10000 distinct -v 2";
print STDERR "Executing: $cmd\n";
`$cmd`;
# count the number of translation in nbest list
$cmd = "wc -l $tmpDir/nbest >> out";
`$cmd`;
`rm -rf $tmpDir`;
}
close(SOURCE);
close(TARGET);
close(ALIGNMENT);
######################
sub Write1Line
{
my ($line, $tmpDir, $fileName) = @_;
open (HANDLE, ">$tmpDir/$fileName");
print HANDLE "$line\n";
close (HANDLE);
}
sub WriteCorpus1Holdout
{
my ($holdoutLineNum, $inFilePath, $tmpDir, $outFileName) = @_;
open (INFILE, "$inFilePath");
open (OUTFILE, ">$tmpDir/$outFileName");
my $lineNum = 0;
while (my $line = <INFILE>) {
chomp($line);
if ($lineNum != $holdoutLineNum) {
print OUTFILE "$line\n";
}
++$lineNum;
}
close (OUTFILE);
close(INFILE);
}

147
contrib/other-builds/extract-mixed-syntax/learnable/reachable.perl
View File

@ -1,147 +0,0 @@
#! /usr/bin/perl -w
use strict;
sub Write1Line;
sub WriteCorpus1Holdout;
my $iniPath = $ARGV[0];
my $isHiero = $ARGV[1];
my $decoderExec = $ARGV[2];
my $extractExec = $ARGV[3];
my $tmpName = $ARGV[4];
my $startLine = $ARGV[5];
my $endLine = $ARGV[6];
print STDERR "iniPath=$iniPath \n isHiero=$isHiero \n decoderExec=$decoderExec \n extractExec=$extractExec \n";
my $WORK_DIR = `pwd`;
chomp($WORK_DIR);
my $MOSES_DIR = "~/workspace/github/mosesdecoder.hieu.gna";
$decoderExec = "$MOSES_DIR/bin/$decoderExec";
$extractExec = "$MOSES_DIR/bin/$extractExec";
my $SPLIT_EXEC = `gsplit --help 2>/dev/null`;
if($SPLIT_EXEC) {
$SPLIT_EXEC = 'gsplit';
}
else {
$SPLIT_EXEC = 'split';
}
my $SORT_EXEC = `gsort --help 2>/dev/null`;
if($SORT_EXEC) {
$SORT_EXEC = 'gsort';
}
else {
$SORT_EXEC = 'sort';
}
my $hieroFlag = "";
if ($isHiero == 1) {
$hieroFlag = "--Hierarchical";
}
print STDERR "WORK_DIR=$WORK_DIR \n";
my $cmd;
open (SOURCE, "source");
open (TARGET, "target");
open (ALIGNMENT, "alignment");
my $numLines = `cat source | wc -l`;
for (my $lineNum = 0; $lineNum < $numLines; ++$lineNum) {
my $source = <SOURCE>; chomp($source);
my $target = <TARGET>; chomp($target);
my $alignment = <ALIGNMENT>; chomp($alignment);
if ($lineNum < $startLine || $lineNum >= $endLine) {
next;
}
#print STDERR "$source ||| $target ||| $alignment \n";
# write out 1 line
my $tmpDir = "$WORK_DIR/$tmpName/work$lineNum";
`mkdir -p $tmpDir`;
Write1Line($source, $tmpDir, "source.1");
Write1Line($target, $tmpDir, "target.1");
Write1Line($alignment, $tmpDir, "alignment.1");
WriteCorpus1Holdout($lineNum, "source", $tmpDir, "source.corpus");
WriteCorpus1Holdout($lineNum, "target", $tmpDir, "target.corpus");
WriteCorpus1Holdout($lineNum, "alignment", $tmpDir, "alignment.corpus");
# train
if ($isHiero == 1) {
$cmd = "$extractExec $tmpDir/target.corpus $tmpDir/source.corpus $tmpDir/alignment.corpus $tmpDir/extract --GZOutput";
}
else {
# pb
$cmd = "$extractExec $tmpDir/target.corpus $tmpDir/source.corpus $tmpDir/alignment.corpus $tmpDir/extract 7 --GZOutput";
}
$cmd = "$MOSES_DIR/scripts/generic/extract-parallel.perl 1 $SPLIT_EXEC $SORT_EXEC $cmd";
print STDERR "Executing: $cmd\n";
`$cmd`;
$cmd = "$MOSES_DIR/scripts/generic/score-parallel.perl 1 $SORT_EXEC $MOSES_DIR/bin/score $tmpDir/extract.sorted.gz /dev/null $tmpDir/pt.half.gz $hieroFlag --NoLex 1";
`$cmd`;
$cmd = "$MOSES_DIR/scripts/generic/score-parallel.perl 1 $SORT_EXEC $MOSES_DIR/bin/score $tmpDir/extract.inv.sorted.gz /dev/null $tmpDir/pt.half.inv.gz --Inverse $hieroFlag --NoLex 1";
`$cmd`;
$cmd = "$MOSES_DIR/bin/consolidate $tmpDir/pt.half.gz $tmpDir/pt.half.inv.gz $tmpDir/pt $hieroFlag --OnlyDirect";
`$cmd`;
# decode
$cmd = "$decoderExec -f $iniPath -feature-overwrite \"TranslationModel0 path=$tmpDir/pt\" -i $tmpDir/source.1 -feature-add \"ConstrainedDecoding path=$tmpDir/target.1\" -v 2";
print STDERR "Executing: $cmd\n";
`$cmd`;
`rm -rf $tmpDir`;
}
close(SOURCE);
close(TARGET);
close(ALIGNMENT);
######################
sub Write1Line
{
my ($line, $tmpDir, $fileName) = @_;
open (HANDLE, ">$tmpDir/$fileName");
print HANDLE "$line\n";
close (HANDLE);
}
sub WriteCorpus1Holdout
{
my ($holdoutLineNum, $inFilePath, $tmpDir, $outFileName) = @_;
open (INFILE, "$inFilePath");
open (OUTFILE, ">$tmpDir/$outFileName");
my $lineNum = 0;
while (my $line = <INFILE>) {
chomp($line);
if ($lineNum != $holdoutLineNum) {
print OUTFILE "$line\n";
}
++$lineNum;
}
close (OUTFILE);
close(INFILE);
}

17
contrib/other-builds/extract-mixed-syntax/learnable/run-parallel.perl
View File

@ -1,17 +0,0 @@
#! /usr/bin/perl -w
my $iniPath = $ARGV[0];
my $SPLIT_LINES = 200;
my $lineCount = `cat source | wc -l`;
print STDERR "lineCount=$lineCount \n";
for (my $startLine = 0; $startLine < $lineCount; $startLine += $SPLIT_LINES) {
my $endLine = $startLine + $SPLIT_LINES;
my $cmd = "../../scripts/reachable.perl $iniPath 1 moses_chart extract-rules tmp-reachable $startLine $endLine &>out.reachable.$startLine &";
print STDERR "Executing: $cmd \n";
system($cmd);
}

69
contrib/other-builds/extract-mixed-syntax/pugiconfig.hpp
View File

@ -1,69 +0,0 @@
/**
* pugixml parser - version 1.2
* --------------------------------------------------------
* Copyright (C) 2006-2012, by Arseny Kapoulkine (arseny.kapoulkine@gmail.com)
* Report bugs and download new versions at http://pugixml.org/
*
* This library is distributed under the MIT License. See notice at the end
* of this file.
*
* This work is based on the pugxml parser, which is:
* Copyright (C) 2003, by Kristen Wegner (kristen@tima.net)
*/
#ifndef HEADER_PUGICONFIG_HPP
#define HEADER_PUGICONFIG_HPP
// Uncomment this to enable wchar_t mode
// #define PUGIXML_WCHAR_MODE
// Uncomment this to disable XPath
// #define PUGIXML_NO_XPATH
// Uncomment this to disable STL
// #define PUGIXML_NO_STL
// Uncomment this to disable exceptions
// #define PUGIXML_NO_EXCEPTIONS
// Set this to control attributes for public classes/functions, i.e.:
// #define PUGIXML_API __declspec(dllexport) // to export all public symbols from DLL
// #define PUGIXML_CLASS __declspec(dllimport) // to import all classes from DLL
// #define PUGIXML_FUNCTION __fastcall // to set calling conventions to all public functions to fastcall
// In absence of PUGIXML_CLASS/PUGIXML_FUNCTION definitions PUGIXML_API is used instead
// Uncomment this to switch to header-only version
// #define PUGIXML_HEADER_ONLY
// #include "pugixml.cpp"
// Tune these constants to adjust memory-related behavior
// #define PUGIXML_MEMORY_PAGE_SIZE 32768
// #define PUGIXML_MEMORY_OUTPUT_STACK 10240
// #define PUGIXML_MEMORY_XPATH_PAGE_SIZE 4096
#endif
/**
* Copyright (c) 2006-2012 Arseny Kapoulkine
*
* Permission is hereby granted, free of charge, to any person
* obtaining a copy of this software and associated documentation
* files (the "Software"), to deal in the Software without
* restriction, including without limitation the rights to use,
* copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following
* conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
* OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
* HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
* WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/

10250
contrib/other-builds/extract-mixed-syntax/pugixml.cpp
View File

File diff suppressed because it is too large Load Diff

1265
contrib/other-builds/extract-mixed-syntax/pugixml.hpp
View File

File diff suppressed because it is too large Load Diff

1
contrib/other-builds/extractor/.cproject
View File

@ -25,6 +25,7 @@
<option id="gnu.cpp.compiler.exe.debug.option.debugging.level.285958391" name="Debug Level" superClass="gnu.cpp.compiler.exe.debug.option.debugging.level" value="gnu.cpp.compiler.debugging.level.max" valueType="enumerated"/>
<option id="gnu.cpp.compiler.option.include.paths.966722418" name="Include paths (-I)" superClass="gnu.cpp.compiler.option.include.paths" valueType="includePath">
<listOptionValue builtIn="false" value="&quot;${workspace_loc}/../../boost/include&quot;"/>
<listOptionValue builtIn="false" value="&quot;${workspace_loc}/../../&quot;"/>
</option>
<inputType id="cdt.managedbuild.tool.gnu.cpp.compiler.input.1839105433" superClass="cdt.managedbuild.tool.gnu.cpp.compiler.input"/>
</tool>

12
contrib/other-builds/moses/.cproject
View File

@ -79,8 +79,18 @@
<fileInfo id="cdt.managedbuild.config.gnu.exe.debug.656913512.511477442" name="Rand.h" rcbsApplicability="disable" resourcePath="LM/Rand.h" toolsToInvoke=""/>
<fileInfo id="cdt.managedbuild.config.gnu.exe.debug.656913512.790052015" name="IRST.h" rcbsApplicability="disable" resourcePath="LM/IRST.h" toolsToInvoke=""/>
<fileInfo id="cdt.managedbuild.config.gnu.exe.debug.656913512.1211280539" name="DALMWrapper.h" rcbsApplicability="disable" resourcePath="LM/DALMWrapper.h" toolsToInvoke=""/>
<folderInfo id="cdt.managedbuild.config.gnu.exe.debug.656913512.1783697903" name="/" resourcePath="TranslationModel/UG">
<toolChain id="cdt.managedbuild.toolchain.gnu.exe.debug.592606983" name="Linux GCC" superClass="cdt.managedbuild.toolchain.gnu.exe.debug" unusedChildren="">
<tool id="cdt.managedbuild.tool.gnu.archiver.base.1088985490" name="GCC Archiver" superClass="cdt.managedbuild.tool.gnu.archiver.base.1976472988"/>
<tool id="cdt.managedbuild.tool.gnu.cpp.compiler.exe.debug.1662972728" name="GCC C++ Compiler" superClass="cdt.managedbuild.tool.gnu.cpp.compiler.exe.debug.1774992327"/>
<tool id="cdt.managedbuild.tool.gnu.c.compiler.exe.debug.1970700220" name="GCC C Compiler" superClass="cdt.managedbuild.tool.gnu.c.compiler.exe.debug.2126314903"/>
<tool id="cdt.managedbuild.tool.gnu.c.linker.exe.debug.1604609677" name="GCC C Linker" superClass="cdt.managedbuild.tool.gnu.c.linker.exe.debug.1168585173"/>
<tool id="cdt.managedbuild.tool.gnu.cpp.linker.exe.debug.1002997408" name="GCC C++ Linker" superClass="cdt.managedbuild.tool.gnu.cpp.linker.exe.debug.2074660557"/>
<tool id="cdt.managedbuild.tool.gnu.assembler.exe.debug.1785961519" name="GCC Assembler" superClass="cdt.managedbuild.tool.gnu.assembler.exe.debug.933467113"/>
</toolChain>
</folderInfo>
<sourceEntries>
<entry excluding="LM/oxlm|LM/Rand.h|LM/Rand.cpp|TranslationModel/CompactPT|LM/NeuralLMWrapper.cpp|FF/PhraseLengthFeatureTest.cpp|PhraseLengthFeatureTest.cpp|LM/BackwardTest.cpp|LM/BackwardLMState.h|LM/BackwardLMState.cpp|LM/Backward.h|LM/Backward.cpp|FeatureVectorTest.cpp|LM/ParallelBackoff.h|LM/ParallelBackoff.cpp|src/SyntacticLanguageModelState.h|src/SyntacticLanguageModelFiles.h|src/SyntacticLanguageModel.h|src/SyntacticLanguageModel.cpp|src/LM/SRI.h|src/LM/SRI.cpp|src/LM/Rand.h|src/LM/Rand.cpp|src/LM/LDHT.h|src/LM/LDHT.cpp|SyntacticLanguageModelState.h|SyntacticLanguageModelFiles.h|SyntacticLanguageModel.h|SyntacticLanguageModel.cpp|LM/LDHT.h|LM/LDHT.cpp" flags="VALUE_WORKSPACE_PATH|RESOLVED" kind="sourcePath" name=""/>
<entry excluding="TranslationModel/UG|LM/oxlm|LM/Rand.h|LM/Rand.cpp|TranslationModel/CompactPT|LM/NeuralLMWrapper.cpp|FF/PhraseLengthFeatureTest.cpp|PhraseLengthFeatureTest.cpp|LM/BackwardTest.cpp|LM/BackwardLMState.h|LM/BackwardLMState.cpp|LM/Backward.h|LM/Backward.cpp|FeatureVectorTest.cpp|LM/ParallelBackoff.h|LM/ParallelBackoff.cpp|src/SyntacticLanguageModelState.h|src/SyntacticLanguageModelFiles.h|src/SyntacticLanguageModel.h|src/SyntacticLanguageModel.cpp|src/LM/SRI.h|src/LM/SRI.cpp|src/LM/Rand.h|src/LM/Rand.cpp|src/LM/LDHT.h|src/LM/LDHT.cpp|SyntacticLanguageModelState.h|SyntacticLanguageModelFiles.h|SyntacticLanguageModel.h|SyntacticLanguageModel.cpp|LM/LDHT.h|LM/LDHT.cpp" flags="VALUE_WORKSPACE_PATH|RESOLVED" kind="sourcePath" name=""/>
</sourceEntries>
</configuration>
</storageModule>

870
contrib/other-builds/moses/.project
View File

@ -536,6 +536,11 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/PCNTools.h</locationURI>
</link>
<link>
<name>PDTAimp.cpp</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/PDTAimp.cpp</locationURI>
</link>
<link>
<name>PDTAimp.h</name>
<type>1</type>
@ -1261,6 +1266,16 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/FF/SetSourcePhrase.h</locationURI>
</link>
<link>
<name>FF/SkeletonChangeInput.cpp</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/FF/SkeletonChangeInput.cpp</locationURI>
</link>
<link>
<name>FF/SkeletonChangeInput.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/FF/SkeletonChangeInput.h</locationURI>
</link>
<link>
<name>FF/SkeletonStatefulFF.cpp</name>
<type>1</type>
@ -1866,6 +1881,11 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/SkeletonPT.h</locationURI>
</link>
<link>
<name>TranslationModel/UG</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/WordCoocTable.cpp</name>
<type>1</type>
@ -1951,6 +1971,11 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/FF/LexicalReordering/SparseReordering.h</locationURI>
</link>
<link>
<name>FF/OSM-Feature/OSM-Feature</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>FF/OSM-Feature/OpSequenceModel.cpp</name>
<type>1</type>
@ -2551,6 +2576,136 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/Scope3Parser/VarSpanTrieBuilder.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/Jamfile</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/Jamfile</locationURI>
</link>
<link>
<name>TranslationModel/UG/Makefile</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/Makefile</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mmsapt.cpp</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mmsapt.cpp</locationURI>
</link>
<link>
<name>TranslationModel/UG/mmsapt.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mmsapt.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mmsapt_align.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mmsapt_align.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mmsapt_phrase_scorers.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mmsapt_phrase_scorers.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/ptable-lookup.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/ptable-lookup.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_phrase_key.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_phrase_key.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_phrase_scorers.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_phrase_scorers.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_base.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_base.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_coherence.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_coherence.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_lex1.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_lex1.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_logcnt.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_logcnt.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_pbwd.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_pbwd.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_pfwd.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_pfwd.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_provenance.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_provenance.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_rareness.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_rareness.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sapt_pscore_unaligned.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sapt_pscore_unaligned.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/sim-pe.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/sim-pe.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/spe-check-coverage.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/spe-check-coverage.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/spe-check-coverage2.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/spe-check-coverage2.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/try-align.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/try-align.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/util</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/fuzzy-match/Alignments.cpp</name>
<type>1</type>
@ -2626,6 +2781,16 @@
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>FF/OSM-Feature/OSM-Feature/KenOSM.cpp</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/FF/OSM-Feature/KenOSM.cpp</locationURI>
</link>
<link>
<name>FF/OSM-Feature/OSM-Feature/KenOSM.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/FF/OSM-Feature/KenOSM.h</locationURI>
</link>
<link>
<name>TranslationModel/CompactPT/bin/gcc-4.7</name>
<type>2</type>
@ -2636,6 +2801,351 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/CompactPT/bin/pt.log</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/Jamfile</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/Jamfile</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/file_io</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/program_options</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/sampling</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/sorting</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/threading</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/Jamfile</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/Jamfile</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/Makefile</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/Makefile</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/calc-coverage.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/calc-coverage.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/custom-pt.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/custom-pt.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mam2symal.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mam2symal.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mam_verify.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mam_verify.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mmlex-build.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mmlex-build.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mmlex-lookup.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mmlex-lookup.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mtt-build.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mtt-build.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mtt-count-words.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mtt-count-words.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mtt-dump.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mtt-dump.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/mtt.count.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/mtt.count.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/num_read_write.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/num_read_write.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/obsolete</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/symal2mam.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/symal2mam.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/test-dynamic-im-tsa.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/test-dynamic-im-tsa.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/tpt_pickler.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/tpt_pickler.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/tpt_pickler.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/tpt_pickler.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/tpt_tightindex.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/tpt_tightindex.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/tpt_tightindex.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/tpt_tightindex.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/tpt_tokenindex.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/tpt_tokenindex.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/tpt_tokenindex.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/tpt_tokenindex.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/tpt_typedefs.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/tpt_typedefs.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_bitext.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_bitext.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_bitext.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_bitext.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_conll_bottom_up_token.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_conll_bottom_up_token.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_conll_record.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_conll_record.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_conll_record.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_conll_record.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_corpus_token.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_corpus_token.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_corpus_token.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_corpus_token.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_deptree.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_deptree.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_deptree.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_deptree.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_im_tsa.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_im_tsa.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_im_ttrack.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_im_ttrack.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_lexical_phrase_scorer1.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_lexical_phrase_scorer1.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_lexical_phrase_scorer2.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_lexical_phrase_scorer2.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_load_primer.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_load_primer.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_load_primer.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_load_primer.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_mm_2d_table.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_mm_2d_table.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_mm_tsa.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_mm_tsa.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_mm_tsa_tree_iterator.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_mm_tsa_tree_iterator.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_mm_ttrack.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_mm_ttrack.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_mmbitext.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_mmbitext.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_mmbitext.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_mmbitext.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_phrasepair.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_phrasepair.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_phrasepair.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_phrasepair.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_tsa_array_entry.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_tsa_array_entry.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_tsa_array_entry.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_tsa_array_entry.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_tsa_base.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_tsa_base.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_tsa_bitset_cache.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_tsa_bitset_cache.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_tsa_tree_iterator.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_tsa_tree_iterator.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_ttrack_base.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_ttrack_base.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_ttrack_base.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_ttrack_base.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_ttrack_position.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_ttrack_position.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_ttrack_position.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_ttrack_position.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/ug_typedefs.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/ug_typedefs.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/util/Makefile</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/util/Makefile</locationURI>
</link>
<link>
<name>TranslationModel/UG/util/ibm1-align</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/util/ibm1-align</locationURI>
</link>
<link>
<name>TranslationModel/UG/util/ibm1-align.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/util/ibm1-align.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/util/tokenindex.dump.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/util/tokenindex.dump.cc</locationURI>
</link>
<link>
<name>bin/BackwardTest.test/gcc-4.7/release</name>
<type>2</type>
@ -2651,6 +3161,81 @@
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/file_io/ug_stream.cpp</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/file_io/ug_stream.cpp</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/file_io/ug_stream.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/file_io/ug_stream.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/program_options/ug_get_options.cpp</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/program_options/ug_get_options.cpp</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/program_options/ug_get_options.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/program_options/ug_get_options.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/program_options/ug_splice_arglist.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/program_options/ug_splice_arglist.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/program_options/ug_splice_arglist.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/program_options/ug_splice_arglist.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/sampling/Sampling.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/sampling/Sampling.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/sorting/NBestList.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/sorting/NBestList.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/sorting/VectorIndexSorter.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/sorting/VectorIndexSorter.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/threading/ug_thread_safe_counter.cc</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/threading/ug_thread_safe_counter.cc</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/threading/ug_thread_safe_counter.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/threading/ug_thread_safe_counter.h</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/obsolete/ug_bitext_base.h</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/obsolete/ug_bitext_base.h</locationURI>
</link>
<link>
<name>bin/BackwardTest.test/gcc-4.7/release/debug-symbols-on</name>
<type>2</type>
@ -2666,6 +3251,21 @@
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>bin/BackwardTest.test/gcc-4.7/release/debug-symbols-on/link-static</name>
<type>2</type>
@ -2681,6 +3281,21 @@
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>bin/BackwardTest.test/gcc-4.7/release/debug-symbols-on/link-static/threading-multi</name>
<type>2</type>
@ -3306,6 +3921,21 @@
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>bin/BackwardTest.test/gcc-4.7/release/debug-symbols-on/link-static/threading-multi/Backward.o</name>
<type>1</type>
@ -3496,6 +4126,71 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/CompactPT/bin/gcc-4.7/release/debug-symbols-on/link-static/threading-multi/ThrowingFwrite.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmsapt_align.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmsapt_align.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ptable-lookup</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ptable-lookup</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ptable-lookup.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ptable-lookup.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/sim-pe</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/sim-pe</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/sim-pe.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/sim-pe.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage2</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage2</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage2.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/spe-check-coverage2.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/try-align</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/try-align</locationURI>
</link>
<link>
<name>TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/try-align.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/try-align.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi</name>
<type>2</type>
<locationURI>virtual:/virtual</locationURI>
</link>
<link>
<name>bin/gcc-4.7/release/debug-symbols-on/link-static/threading-multi/TranslationModel/CYKPlusParser/ChartRuleLookupManagerCYKPlus.o</name>
<type>1</type>
@ -3651,5 +4346,180 @@
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/bin/gcc-4.7/release/debug-symbols-on/link-static/threading-multi/TranslationModel/fuzzy-match/create_xml.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_get_options.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_get_options.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_splice_arglist.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_splice_arglist.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_stream.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_stream.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_thread_safe_counter.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/generic/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_thread_safe_counter.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/calc-coverage</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/calc-coverage</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/calc-coverage.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/calc-coverage.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam2symal</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam2symal</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam2symal.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam2symal.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam_verify</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam_verify</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam_verify.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mam_verify.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-build</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-build</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-build.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-build.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-lookup</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-lookup</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-lookup.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mmlex-lookup.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-build</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-build</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-build.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-build.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-count-words</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-count-words</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-count-words.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-count-words.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-dump</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-dump</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-dump.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/mtt-dump.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/symal2mam</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/symal2mam</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/symal2mam.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/symal2mam.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/tpt_pickler.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/tpt_pickler.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/tpt_tightindex.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/tpt_tightindex.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/tpt_tokenindex.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/tpt_tokenindex.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_bitext.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_bitext.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_conll_record.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_conll_record.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_corpus_token.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_corpus_token.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_deptree.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_deptree.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_load_primer.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_load_primer.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_mmbitext.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_mmbitext.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_phrasepair.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_phrasepair.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_tsa_array_entry.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_tsa_array_entry.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_ttrack_base.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_ttrack_base.o</locationURI>
</link>
<link>
<name>TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_ttrack_position.o</name>
<type>1</type>
<locationURI>PARENT-3-PROJECT_LOC/moses/TranslationModel/UG/mm/bin/gcc-4.8/release/debug-symbols-on/link-static/threading-multi/ug_ttrack_position.o</locationURI>
</link>
</linkedResources>
</projectDescription>

4
mert/FeatureStats.h
View File

@ -16,9 +16,7 @@
#include <vector>
#include <boost/unordered_map.hpp>
#include <util/string_piece.hh>
#include "util/string_piece.hh"
#include "Types.h"
namespace MosesTuning

30
mert/HopeFearDecoder.cpp
View File

@ -17,6 +17,7 @@ License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
***********************************************************************/
#include <algorithm>
#include <cmath>
#include <iterator>
@ -172,12 +173,14 @@ HypergraphHopeFearDecoder::HypergraphHopeFearDecoder
static const string kWeights = "weights";
fs::directory_iterator dend;
size_t fileCount = 0;
cerr << "Reading hypergraphs" << endl;
cerr << "Reading hypergraphs" << endl;
for (fs::directory_iterator di(hypergraphDir); di != dend; ++di) {
if (di->path().filename() == kWeights) continue;
const fs::path& hgpath = di->path();
if (hgpath.filename() == kWeights) continue;
Graph graph(vocab_);
size_t id = boost::lexical_cast<size_t>(di->path().stem().string());
util::scoped_fd fd(util::OpenReadOrThrow(di->path().string().c_str()));
size_t id = boost::lexical_cast<size_t>(hgpath.stem().string());
util::scoped_fd fd(util::OpenReadOrThrow(hgpath.string().c_str()));
//util::FilePiece file(di->path().string().c_str());
util::FilePiece file(fd.release());
ReadGraph(file,graph);
@ -195,19 +198,24 @@ HypergraphHopeFearDecoder::HypergraphHopeFearDecoder
}
cerr << endl << "Done" << endl;
sentenceIds_.resize(graphs_.size());
for (size_t i = 0; i < graphs_.size(); ++i) sentenceIds_[i] = i;
if (!no_shuffle) {
random_shuffle(sentenceIds_.begin(), sentenceIds_.end());
}
}
void HypergraphHopeFearDecoder::reset() {
graphIter_ = graphs_.begin();
sentenceIdIter_ = sentenceIds_.begin();
}
void HypergraphHopeFearDecoder::next() {
++graphIter_;
sentenceIdIter_++;
}
bool HypergraphHopeFearDecoder::finished() {
return graphIter_ == graphs_.end();
return sentenceIdIter_ == sentenceIds_.end();
}
void HypergraphHopeFearDecoder::HopeFear(
@ -215,10 +223,10 @@ void HypergraphHopeFearDecoder::HopeFear(
const MiraWeightVector& wv,
HopeFearData* hopeFear
) {
size_t sentenceId = graphIter_->first;
size_t sentenceId = *sentenceIdIter_;
SparseVector weights;
wv.ToSparse(&weights);
const Graph& graph = *(graphIter_->second);
const Graph& graph = *(graphs_[sentenceId]);
ValType hope_scale = 1.0;
HgHypothesis hopeHypo, fearHypo, modelHypo;
@ -309,11 +317,11 @@ void HypergraphHopeFearDecoder::HopeFear(
void HypergraphHopeFearDecoder::MaxModel(const AvgWeightVector& wv, vector<ValType>* stats) {
assert(!finished());
HgHypothesis bestHypo;
size_t sentenceId = graphIter_->first;
size_t sentenceId = *sentenceIdIter_;
SparseVector weights;
wv.ToSparse(&weights);
vector<ValType> bg(kBleuNgramOrder*2+1);
Viterbi(*(graphIter_->second), weights, 0, references_, sentenceId, bg, &bestHypo);
Viterbi(*(graphs_[sentenceId]), weights, 0, references_, sentenceId, bg, &bestHypo);
stats->resize(bestHypo.bleuStats.size());
/*
for (size_t i = 0; i < bestHypo.text.size(); ++i) {

3
mert/HopeFearDecoder.h
View File

@ -140,7 +140,8 @@ private:
//maps sentence Id to graph ptr
typedef std::map<size_t, boost::shared_ptr<Graph> > GraphColl;
GraphColl graphs_;
GraphColl::const_iterator graphIter_;
std::vector<size_t> sentenceIds_;
std::vector<size_t>::const_iterator sentenceIdIter_;
ReferenceSet references_;
Vocab vocab_;
};

3
moses-chart-cmd/Main.cpp
View File

@ -301,6 +301,9 @@ int main(int argc, char* argv[])
while(ReadInput(*ioWrapper,staticData.GetInputType(),source)) {
IFVERBOSE(1)
ResetUserTime();
FeatureFunction::CallChangeSource(source);
TranslationTask *task = new TranslationTask(source, *ioWrapper);
source = NULL; // task will delete source
#ifdef WITH_THREADS

3
moses-cmd/Main.cpp
View File

@ -758,6 +758,9 @@ int main(int argc, char** argv)
IFVERBOSE(1) {
ResetUserTime();
}
FeatureFunction::CallChangeSource(source);
// set up task of translating one sentence
TranslationTask* task =
new TranslationTask(lineCount,source, outputCollector.get(),

4
moses/ChartParser.cpp
View File

@ -85,7 +85,7 @@ void ChartParserUnknown::Process(const Word &sourceWord, const WordsRange &range
UTIL_THROW_IF2(targetLHS->GetFactor(0) == NULL, "Null factor for target LHS");
// add to dictionary
TargetPhrase *targetPhrase = new TargetPhrase();
TargetPhrase *targetPhrase = new TargetPhrase(NULL);
Word &targetWord = targetPhrase->AddWord();
targetWord.CreateUnknownWord(sourceWord);
@ -108,7 +108,7 @@ void ChartParserUnknown::Process(const Word &sourceWord, const WordsRange &range
// drop source word. create blank trans opt
float unknownScore = FloorScore(-numeric_limits<float>::infinity());
TargetPhrase *targetPhrase = new TargetPhrase();
TargetPhrase *targetPhrase = new TargetPhrase(NULL);
// loop
const UnknownLHSList &lhsList = staticData.GetUnknownLHS();
UnknownLHSList::const_iterator iterLHS;

14
moses/DecodeGraph.h
View File

@ -38,7 +38,7 @@ class DecodeGraph
{
protected:
std::list<const DecodeStep*> m_steps;
size_t m_position;
size_t m_id; // contiguous unique id, starting from 0
size_t m_maxChartSpan;
size_t m_backoff;
@ -46,15 +46,15 @@ public:
/**
* position: The position of this graph within the decode sequence.
**/
DecodeGraph(size_t position)
: m_position(position)
DecodeGraph(size_t id)
: m_id(id)
, m_maxChartSpan(NOT_FOUND)
, m_backoff(0)
{}
// for chart decoding
DecodeGraph(size_t position, size_t maxChartSpan)
: m_position(position)
DecodeGraph(size_t id, size_t maxChartSpan)
: m_id(id)
, m_maxChartSpan(maxChartSpan) {
}
@ -90,8 +90,8 @@ public:
m_backoff = backoff;
}
size_t GetPosition() const {
return m_position;
size_t GetId() const {
return m_id;
}
};

2
moses/FF/Factory.cpp
View File

@ -48,6 +48,7 @@
#include "NieceTerminal.h"
#include "SpanLength.h"
#include "SyntaxRHS.h"
#include "SkeletonChangeInput.h"
#include "moses/FF/SkeletonStatelessFF.h"
#include "moses/FF/SkeletonStatefulFF.h"
@ -208,6 +209,7 @@ FeatureRegistry::FeatureRegistry()
MOSES_FNAME(SparseHieroReorderingFeature);
MOSES_FNAME(SpanLength);
MOSES_FNAME(SyntaxRHS);
MOSES_FNAME(SkeletonChangeInput);
MOSES_FNAME(SkeletonStatelessFF);
MOSES_FNAME(SkeletonStatefulFF);

8
moses/FF/FeatureFunction.cpp
View File

@ -34,6 +34,14 @@ void FeatureFunction::Destroy()
RemoveAllInColl(s_staticColl);
}
void FeatureFunction::CallChangeSource(InputType *&input)
{
for (size_t i = 0; i < s_staticColl.size(); ++i) {
const FeatureFunction &ff = *s_staticColl[i];
ff.ChangeSource(input);
}
}
FeatureFunction::
FeatureFunction(const std::string& line)
: m_tuneable(true)

6
moses/FF/FeatureFunction.h
View File

@ -48,6 +48,8 @@ public:
static FeatureFunction &FindFeatureFunction(const std::string& name);
static void Destroy();
static void CallChangeSource(InputType *&input);
FeatureFunction(const std::string &line);
FeatureFunction(size_t numScoreComponents, const std::string &line);
virtual bool IsStateless() const = 0;
@ -109,6 +111,10 @@ public:
, ScoreComponentCollection &scoreBreakdown
, ScoreComponentCollection &estimatedFutureScore) const = 0;
// override this method if you want to change the input before decoding
virtual void ChangeSource(InputType *&input) const
{}
// This method is called once all the translation options are retrieved from the phrase table, and
// just before search.
// 'inputPath' is guaranteed to be the raw substring from the input. No factors were added or taken away

1
moses/FF/OSM-Feature/OpSequenceModel.cpp
View File

@ -25,7 +25,6 @@ OpSequenceModel::~OpSequenceModel() {
void OpSequenceModel :: readLanguageModel(const char *lmFile)
{
string unkOp = "_TRANS_SLF_";
OSM = ConstructOSMLM(m_lmPath);

37
moses/FF/PhrasePenalty.cpp
View File

@ -1,11 +1,16 @@
#include <vector>
#include "PhrasePenalty.h"
#include "moses/ScoreComponentCollection.h"
#include "moses/TranslationModel/PhraseDictionary.h"
#include "util/exception.hh"
using namespace std;
namespace Moses
{
PhrasePenalty::PhrasePenalty(const std::string &line)
: StatelessFeatureFunction(1, line)
: StatelessFeatureFunction(1, line)
, m_perPhraseTable(false)
{
ReadParameters();
}
@ -15,8 +20,34 @@ void PhrasePenalty::EvaluateInIsolation(const Phrase &source
, ScoreComponentCollection &scoreBreakdown
, ScoreComponentCollection &estimatedFutureScore) const
{
scoreBreakdown.Assign(this, 1.0f);
if (m_perPhraseTable) {
const PhraseDictionary *pt = targetPhrase.GetContainer();
if (pt) {
size_t ptId = pt->GetId();
UTIL_THROW_IF2(ptId >= m_numScoreComponents, "Wrong number of scores");
vector<float> scores(m_numScoreComponents, 0);
scores[ptId] = 1.0f;
scoreBreakdown.Assign(this, scores);
}
}
else {
scoreBreakdown.Assign(this, 1.0f);
}
}
void PhrasePenalty::SetParameter(const std::string& key, const std::string& value)
{
if (key == "per-phrase-table") {
m_perPhraseTable =Scan<bool>(value);
}
else {
StatelessFeatureFunction::SetParameter(key, value);
}
}
} // namespace

4
moses/FF/PhrasePenalty.h
View File

@ -34,6 +34,10 @@ public:
, ScoreComponentCollection *estimatedFutureScore = NULL) const
{}
void SetParameter(const std::string& key, const std::string& value);
protected:
bool m_perPhraseTable;
};
} //namespace

92
moses/FF/SkeletonChangeInput.cpp Normal file
View File

@ -0,0 +1,92 @@
#include <vector>
#include "SkeletonChangeInput.h"
#include "moses/ScoreComponentCollection.h"
#include "moses/TargetPhrase.h"
#include "moses/Sentence.h"
#include "moses/FactorCollection.h"
#include "util/exception.hh"
using namespace std;
namespace Moses
{
SkeletonChangeInput::SkeletonChangeInput(const std::string &line)
:StatelessFeatureFunction(2, line)
{
ReadParameters();
}
void SkeletonChangeInput::EvaluateInIsolation(const Phrase &source
, const TargetPhrase &targetPhrase
, ScoreComponentCollection &scoreBreakdown
, ScoreComponentCollection &estimatedFutureScore) const
{
// dense scores
vector<float> newScores(m_numScoreComponents);
newScores[0] = 1.5;
newScores[1] = 0.3;
scoreBreakdown.PlusEquals(this, newScores);
// sparse scores
scoreBreakdown.PlusEquals(this, "sparse-name", 2.4);
}
void SkeletonChangeInput::EvaluateWithSourceContext(const InputType &input
, const InputPath &inputPath
, const TargetPhrase &targetPhrase
, const StackVec *stackVec
, ScoreComponentCollection &scoreBreakdown
, ScoreComponentCollection *estimatedFutureScore) const
{
if (targetPhrase.GetNumNonTerminals()) {
vector<float> newScores(m_numScoreComponents);
newScores[0] = - std::numeric_limits<float>::infinity();
scoreBreakdown.PlusEquals(this, newScores);
}
}
void SkeletonChangeInput::EvaluateWhenApplied(const Hypothesis& hypo,
ScoreComponentCollection* accumulator) const
{}
void SkeletonChangeInput::EvaluateWhenApplied(const ChartHypothesis &hypo,
ScoreComponentCollection* accumulator) const
{}
void SkeletonChangeInput::ChangeSource(InputType *&input) const
{
// add factor[1] to each word. Created from first 4 letter of factor[0]
Sentence *sentence = dynamic_cast<Sentence*>(input);
UTIL_THROW_IF2(sentence == NULL, "Not a sentence input");
FactorCollection &fc = FactorCollection::Instance();
size_t size = sentence->GetSize();
for (size_t i = 0; i < size; ++i) {
Word &word = sentence->Phrase::GetWord(i);
const Factor *factor0 = word[0];
std::string str = factor0->GetString().as_string();
if (str.length() > 4) {
str = str.substr(0, 4);
}
const Factor *factor1 = fc.AddFactor(str);
word.SetFactor(1, factor1);
}
}
void SkeletonChangeInput::SetParameter(const std::string& key, const std::string& value)
{
if (key == "arg") {
// set value here
} else {
StatelessFeatureFunction::SetParameter(key, value);
}
}
}

41
moses/FF/SkeletonChangeInput.h Normal file
View File

@ -0,0 +1,41 @@
#pragma once
#include <string>
#include "StatelessFeatureFunction.h"
namespace Moses
{
class SkeletonChangeInput : public StatelessFeatureFunction
{
public:
SkeletonChangeInput(const std::string &line);
bool IsUseable(const FactorMask &mask) const {
return true;
}
void EvaluateInIsolation(const Phrase &source
, const TargetPhrase &targetPhrase
, ScoreComponentCollection &scoreBreakdown
, ScoreComponentCollection &estimatedFutureScore) const;
void ChangeSource(InputType *&input) const;
void EvaluateWithSourceContext(const InputType &input
, const InputPath &inputPath
, const TargetPhrase &targetPhrase
, const StackVec *stackVec
, ScoreComponentCollection &scoreBreakdown
, ScoreComponentCollection *estimatedFutureScore = NULL) const;
void EvaluateWhenApplied(const Hypothesis& hypo,
ScoreComponentCollection* accumulator) const;
void EvaluateWhenApplied(const ChartHypothesis &hypo,
ScoreComponentCollection* accumulator) const;
void SetParameter(const std::string& key, const std::string& value);
};
}

14
moses/Hypothesis.cpp
View File

@ -85,14 +85,13 @@ Hypothesis::Hypothesis(const Hypothesis &prevHypo, const TranslationOption &tran
, m_wordDeleted(false)
, m_totalScore(0.0f)
, m_futureScore(0.0f)
, m_scoreBreakdown(prevHypo.GetScoreBreakdown())
, m_ffStates(prevHypo.m_ffStates.size())
, m_arcList(NULL)
, m_transOpt(transOpt)
, m_manager(prevHypo.GetManager())
, m_id(m_manager.GetNextHypoId())
{
m_scoreBreakdown.PlusEquals(transOpt.GetScoreBreakdown());
m_currScoreBreakdown.PlusEquals(transOpt.GetScoreBreakdown());
// assert that we are not extending our hypothesis by retranslating something
// that this hypothesis has already translated!
@ -214,7 +213,7 @@ void Hypothesis::EvaluateWith(const StatefulFeatureFunction &sfff,
m_ffStates[state_idx] = sfff.EvaluateWhenApplied(
*this,
m_prevHypo ? m_prevHypo->m_ffStates[state_idx] : NULL,
&m_scoreBreakdown);
&m_currScoreBreakdown);
}
}
@ -222,7 +221,7 @@ void Hypothesis::EvaluateWith(const StatelessFeatureFunction& slff)
{
const StaticData &staticData = StaticData::Instance();
if (! staticData.IsFeatureFunctionIgnored( slff )) {
slff.EvaluateWhenApplied(*this, &m_scoreBreakdown);
slff.EvaluateWhenApplied(*this, &m_currScoreBreakdown);
}
}
@ -256,7 +255,7 @@ void Hypothesis::Evaluate(const SquareMatrix &futureScore)
if (! staticData.IsFeatureFunctionIgnored(ff)) {
m_ffStates[i] = ff.EvaluateWhenApplied(*this,
m_prevHypo ? m_prevHypo->m_ffStates[i] : NULL,
&m_scoreBreakdown);
&m_currScoreBreakdown);
}
}
@ -269,7 +268,8 @@ void Hypothesis::Evaluate(const SquareMatrix &futureScore)
m_futureScore = futureScore.CalcFutureScore( m_sourceCompleted );
// TOTAL
m_totalScore = m_scoreBreakdown.GetWeightedScore() + m_futureScore;
m_totalScore = m_currScoreBreakdown.GetWeightedScore() + m_futureScore;
if (m_prevHypo) m_totalScore += m_prevHypo->GetScore();
IFVERBOSE(2) {
m_manager.GetSentenceStats().StopTimeEstimateScore();
@ -315,7 +315,7 @@ void Hypothesis::PrintHypothesis() const
// TRACE_ERR( "\tlanguage model cost "); // <<m_score[ScoreType::LanguageModelScore]<<endl;
// TRACE_ERR( "\tword penalty "); // <<(m_score[ScoreType::WordPenalty]*weightWordPenalty)<<endl;
TRACE_ERR( "\tscore "<<m_totalScore - m_futureScore<<" + future cost "<<m_futureScore<<" = "<<m_totalScore<<endl);
TRACE_ERR( "\tunweighted feature scores: " << m_scoreBreakdown << endl);
TRACE_ERR( "\tunweighted feature scores: " << m_currScoreBreakdown << endl);
//PrintLMScores();
}

16
moses/Hypothesis.h
View File

@ -25,6 +25,9 @@ Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
#include <iostream>
#include <memory>
#include <boost/scoped_ptr.hpp>
#include <vector>
#include "Phrase.h"
#include "TypeDef.h"
@ -77,7 +80,9 @@ protected:
bool m_wordDeleted;
float m_totalScore; /*! score so far */
float m_futureScore; /*! estimated future cost to translate rest of sentence */
ScoreComponentCollection m_scoreBreakdown; /*! scores for this hypothesis */
/*! sum of scores of this hypothesis, and previous hypotheses. Lazily initialised. */
mutable boost::scoped_ptr<ScoreComponentCollection> m_scoreBreakdown;
ScoreComponentCollection m_currScoreBreakdown; /*! scores for this hypothesis only */
std::vector<const FFState*> m_ffStates;
const Hypothesis *m_winningHypo;
ArcList *m_arcList; /*! all arcs that end at the same trellis point as this hypothesis */
@ -228,7 +233,14 @@ public:
return m_arcList;
}
const ScoreComponentCollection& GetScoreBreakdown() const {
return m_scoreBreakdown;
if (!m_scoreBreakdown.get()) {
m_scoreBreakdown.reset(new ScoreComponentCollection());
m_scoreBreakdown->PlusEquals(m_currScoreBreakdown);
if (m_prevHypo) {
m_scoreBreakdown->PlusEquals(m_prevHypo->GetScoreBreakdown());
}
}
return *(m_scoreBreakdown.get());
}
float GetTotalScore() const {
return m_totalScore;

2
moses/MockHypothesis.cpp
View File

@ -62,7 +62,7 @@ MockHypothesisGuard::MockHypothesisGuard(
for (; ti != targetSegments.end() && ai != alignments.end(); ++ti,++ai) {
Hypothesis* prevHypo = m_hypothesis;
WordsRange wordsRange(ai->first,ai->second);
m_targetPhrases.push_back(TargetPhrase());
m_targetPhrases.push_back(TargetPhrase(NULL));
// m_targetPhrases.back().CreateFromString(Input, factors, *ti, "|", NULL);
m_targetPhrases.back().CreateFromString(Input, factors, *ti, NULL);
m_toptions.push_back(new TranslationOption

463
moses/PDTAimp.cpp Normal file
View File

@ -0,0 +1,463 @@
#include "PDTAimp.h"
namespace Moses
{
PDTAimp::PDTAimp(PhraseDictionaryTreeAdaptor *p)
: m_dict(0),
m_obj(p),
useCache(1),
totalE(0),
distinctE(0) {
m_numInputScores = 0;
const StaticData &staticData = StaticData::Instance();
m_inputFeature = &InputFeature::Instance();
if (m_inputFeature) {
const PhraseDictionary *firstPt = PhraseDictionary::GetColl()[0];
if (firstPt == m_obj) {
m_numInputScores = m_inputFeature->GetNumScoreComponents();
}
}
}
PDTAimp::~PDTAimp() {
CleanUp();
delete m_dict;
if (StaticData::Instance().GetVerboseLevel() >= 2) {
TRACE_ERR("tgt candidates stats: total="<<totalE<<"; distinct="
<<distinctE<<" ("<<distinctE/(0.01*totalE)<<"); duplicates="
<<totalE-distinctE<<" ("<<(totalE-distinctE)/(0.01*totalE)
<<")\n");
TRACE_ERR("\npath statistics\n");
if(path1Best.size()) {
TRACE_ERR("1-best: ");
std::copy(path1Best.begin()+1,path1Best.end(),
std::ostream_iterator<size_t>(std::cerr," \t"));
TRACE_ERR("\n");
}
if(pathCN.size()) {
TRACE_ERR("CN (full): ");
std::transform(pathCN.begin()+1
,pathCN.end()
,std::ostream_iterator<double>(std::cerr," \t")
,Exp);
TRACE_ERR("\n");
}
if(pathExplored.size()) {
TRACE_ERR("CN (explored): ");
std::copy(pathExplored.begin()+1,pathExplored.end(),
std::ostream_iterator<size_t>(std::cerr," \t"));
TRACE_ERR("\n");
}
}
}
void PDTAimp::CleanUp() {
assert(m_dict);
m_dict->FreeMemory();
for(size_t i=0; i<m_tgtColls.size(); ++i) delete m_tgtColls[i];
m_tgtColls.clear();
m_cache.clear();
m_rangeCache.clear();
uniqSrcPhr.clear();
}
TargetPhraseCollectionWithSourcePhrase const*
PDTAimp::GetTargetPhraseCollection(Phrase const &src) const {
assert(m_dict);
if(src.GetSize()==0) return 0;
std::pair<MapSrc2Tgt::iterator,bool> piter;
if(useCache) {
piter=m_cache.insert(std::make_pair(src,static_cast<TargetPhraseCollectionWithSourcePhrase const*>(0)));
if(!piter.second) return piter.first->second;
} else if (m_cache.size()) {
MapSrc2Tgt::const_iterator i=m_cache.find(src);
return (i!=m_cache.end() ? i->second : 0);
}
std::vector<std::string> srcString(src.GetSize());
// convert source Phrase into vector of strings
for(size_t i=0; i<srcString.size(); ++i) {
Factors2String(src.GetWord(i),srcString[i]);
}
// get target phrases in string representation
std::vector<StringTgtCand> cands;
std::vector<std::string> wacands;
m_dict->GetTargetCandidates(srcString,cands,wacands);
if(cands.empty()) {
return 0;
}
//TODO: Multiple models broken here
std::vector<float> weights = StaticData::Instance().GetWeights(m_obj);
std::vector<TargetPhrase> tCands;
tCands.reserve(cands.size());
std::vector<std::pair<float,size_t> > costs;
costs.reserve(cands.size());
std::vector<Phrase> sourcePhrases;
sourcePhrases.reserve(cands.size());
// convert into TargetPhrases
for(size_t i=0; i<cands.size(); ++i) {
TargetPhrase targetPhrase(m_obj);
StringTgtCand::Tokens const& factorStrings=cands[i].tokens;
Scores const& probVector=cands[i].scores;
std::vector<float> scoreVector(probVector.size());
std::transform(probVector.begin(),probVector.end(),scoreVector.begin(),
TransformScore);
std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),
FloorScore);
//sparse features.
//These are already in log-space
for (size_t j = 0; j < cands[i].fnames.size(); ++j) {
targetPhrase.GetScoreBreakdown().Assign(m_obj, *cands[i].fnames[j], cands[i].fvalues[j]);
}
CreateTargetPhrase(targetPhrase,factorStrings,scoreVector, Scores(0), &wacands[i], &src);
costs.push_back(std::make_pair(-targetPhrase.GetFutureScore(),tCands.size()));
tCands.push_back(targetPhrase);
sourcePhrases.push_back(src);
}
TargetPhraseCollectionWithSourcePhrase *rv;
rv=PruneTargetCandidates(tCands,costs, sourcePhrases);
if(rv->IsEmpty()) {
delete rv;
return 0;
} else {
if(useCache) piter.first->second=rv;
m_tgtColls.push_back(rv);
return rv;
}
}
void PDTAimp::Create(const std::vector<FactorType> &input
, const std::vector<FactorType> &output
, const std::string &filePath
, const std::vector<float> &weight
) {
// set my members
m_dict=new PhraseDictionaryTree();
m_input=input;
m_output=output;
const StaticData &staticData = StaticData::Instance();
m_dict->NeedAlignmentInfo(staticData.NeedAlignmentInfo());
std::string binFname=filePath+".binphr.idx";
if(!FileExists(binFname.c_str())) {
UTIL_THROW2( "bin ttable does not exist");
//TRACE_ERR( "bin ttable does not exist -> create it\n");
//InputFileStream in(filePath);
//m_dict->Create(in,filePath);
}
VERBOSE(1,"reading bin ttable\n");
// m_dict->Read(filePath);
bool res=m_dict->Read(filePath);
if (!res) {
std::stringstream strme;
strme << "bin ttable was read in a wrong way\n";
UserMessage::Add(strme.str());
exit(1);
}
}
void PDTAimp::CacheSource(ConfusionNet const& src) {
assert(m_dict);
const size_t srcSize=src.GetSize();
std::vector<size_t> exploredPaths(srcSize+1,0);
std::vector<double> exPathsD(srcSize+1,-1.0);
// collect some statistics
std::vector<size_t> cnDepths(srcSize,0);
for(size_t i=0; i<srcSize; ++i) cnDepths[i]=src[i].size();
for(size_t len=1; len<=srcSize; ++len)
for(size_t i=0; i<=srcSize-len; ++i) {
double pd=0.0;
for(size_t k=i; k<i+len; ++k) pd+=log(1.0*cnDepths[k]);
exPathsD[len]=(exPathsD[len]>=0.0 ? addLogScale(pd,exPathsD[len]) : pd);
}
// update global statistics
if(pathCN.size()<=srcSize) pathCN.resize(srcSize+1,-1.0);
for(size_t len=1; len<=srcSize; ++len)
pathCN[len]=pathCN[len]>=0.0 ? addLogScale(pathCN[len],exPathsD[len]) : exPathsD[len];
if(path1Best.size()<=srcSize) path1Best.resize(srcSize+1,0);
for(size_t len=1; len<=srcSize; ++len) path1Best[len]+=srcSize-len+1;
if (StaticData::Instance().GetVerboseLevel() >= 2 && exPathsD.size()) {
TRACE_ERR("path stats for current CN: \nCN (full): ");
std::transform(exPathsD.begin()+1
,exPathsD.end()
,std::ostream_iterator<double>(std::cerr," ")
,Exp);
TRACE_ERR("\n");
}
typedef StringTgtCand::Tokens sPhrase;
typedef std::map<StringTgtCand::Tokens,TScores> E2Costs;
std::map<Range,E2Costs> cov2cand;
std::vector<State> stack;
for(Position i=0 ; i < srcSize ; ++i)
stack.push_back(State(i, i, m_dict->GetRoot(), std::vector<float>(m_numInputScores,0.0)));
std::vector<float> weightTrans = StaticData::Instance().GetWeights(m_obj);
std::vector<float> weightInput = StaticData::Instance().GetWeights(m_inputFeature);
float weightWP = StaticData::Instance().GetWeightWordPenalty();
while(!stack.empty()) {
State curr(stack.back());
stack.pop_back();
UTIL_THROW_IF2(curr.end() >= srcSize, "Error");
const ConfusionNet::Column &currCol=src[curr.end()];
// in a given column, loop over all possibilities
for(size_t colidx=0; colidx<currCol.size(); ++colidx) {
const Word& w=currCol[colidx].first; // w=the i^th possibility in column colidx
std::string s;
Factors2String(w,s);
bool isEpsilon=(s=="" || s==EPSILON);
//assert that we have the right number of link params in this CN option
UTIL_THROW_IF2(currCol[colidx].second.denseScores.size() < m_numInputScores,
"Incorrect number of input scores");
// do not start with epsilon (except at first position)
if(isEpsilon && curr.begin()==curr.end() && curr.begin()>0) continue;
// At a given node in the prefix tree, look to see if w defines an edge to
// another node (Extend). Stay at the same node if w==EPSILON
PPtr nextP = (isEpsilon ? curr.ptr : m_dict->Extend(curr.ptr,s));
if(nextP) { // w is a word that should be considered
Range newRange(curr.begin(),curr.end()+src.GetColumnIncrement(curr.end(),colidx));
//add together the link scores from the current state and the new arc
float inputScoreSum = 0;
std::vector<float> newInputScores(m_numInputScores,0.0);
if (m_numInputScores) {
std::transform(currCol[colidx].second.denseScores.begin(), currCol[colidx].second.denseScores.end(),
curr.GetScores().begin(),
newInputScores.begin(),
std::plus<float>());
//we need to sum up link weights (excluding realWordCount, which isn't in numLinkParams)
//if the sum is too low, then we won't expand this.
//TODO: dodgy! shouldn't we consider weights here? what about zero-weight params?
inputScoreSum = std::accumulate(newInputScores.begin(),newInputScores.begin()+m_numInputScores,0.0);
}
Phrase newSrc(curr.src);
if(!isEpsilon) newSrc.AddWord(w);
if(newRange.second<srcSize && inputScoreSum>LOWEST_SCORE) {
// if there is more room to grow, add a new state onto the queue
// to be explored that represents [begin, curEnd+)
stack.push_back(State(newRange,nextP,newInputScores));
stack.back().src=newSrc;
}
std::vector<StringTgtCand> tcands;
// now, look up the target candidates (aprx. TargetPhraseCollection) for
// the current path through the CN
m_dict->GetTargetCandidates(nextP,tcands);
if(newRange.second>=exploredPaths.size()+newRange.first)
exploredPaths.resize(newRange.second-newRange.first+1,0);
++exploredPaths[newRange.second-newRange.first];
totalE+=tcands.size();
if(tcands.size()) {
E2Costs& e2costs=cov2cand[newRange];
Phrase const* srcPtr=uniqSrcPhr(newSrc);
for(size_t i=0; i<tcands.size(); ++i) {
//put input scores in first - already logged, just drop in directly
std::vector<float> transcores(m_obj->GetNumScoreComponents());
UTIL_THROW_IF2(transcores.size() != weightTrans.size(),
"Incorrect number of translation scores");
//put in phrase table scores, logging as we insert
std::transform(tcands[i].scores.begin()
,tcands[i].scores.end()
,transcores.begin()
,TransformScore);
//tally up
float score=std::inner_product(transcores.begin(), transcores.end(), weightTrans.begin(), 0.0f);
// input feature
score +=std::inner_product(newInputScores.begin(), newInputScores.end(), weightInput.begin(), 0.0f);
//count word penalty
score-=tcands[i].tokens.size() * weightWP;
std::pair<E2Costs::iterator,bool> p=e2costs.insert(std::make_pair(tcands[i].tokens,TScores()));
if(p.second) ++distinctE;
TScores & scores=p.first->second;
if(p.second || scores.total<score) {
scores.total=score;
scores.transScore=transcores;
scores.inputScores=newInputScores;
scores.src=srcPtr;
}
}
}
}
}
} // end while(!stack.empty())
if (StaticData::Instance().GetVerboseLevel() >= 2 && exploredPaths.size()) {
TRACE_ERR("CN (explored): ");
std::copy(exploredPaths.begin()+1,exploredPaths.end(),
std::ostream_iterator<size_t>(std::cerr," "));
TRACE_ERR("\n");
}
if(pathExplored.size()<exploredPaths.size())
pathExplored.resize(exploredPaths.size(),0);
for(size_t len=1; len<=srcSize; ++len)
pathExplored[len]+=exploredPaths[len];
m_rangeCache.resize(src.GetSize(),vTPC(src.GetSize(),0));
for(std::map<Range,E2Costs>::const_iterator i=cov2cand.begin(); i!=cov2cand.end(); ++i) {
assert(i->first.first<m_rangeCache.size());
assert(i->first.second>0);
assert(static_cast<size_t>(i->first.second-1)<m_rangeCache[i->first.first].size());
assert(m_rangeCache[i->first.first][i->first.second-1]==0);
std::vector<TargetPhrase> tCands;
tCands.reserve(i->second.size());
std::vector<std::pair<float,size_t> > costs;
costs.reserve(i->second.size());
std::vector<Phrase> sourcePhrases;
sourcePhrases.reserve(i->second.size());
for(E2Costs::const_iterator j=i->second.begin(); j!=i->second.end(); ++j) {
TScores const & scores=j->second;
TargetPhrase targetPhrase(m_obj);
CreateTargetPhrase(targetPhrase
, j ->first
, scores.transScore
, scores.inputScores
, NULL
, scores.src);
costs.push_back(std::make_pair(-targetPhrase.GetFutureScore(),tCands.size()));
tCands.push_back(targetPhrase);
sourcePhrases.push_back(*scores.src);
//std::cerr << i->first.first << "-" << i->first.second << ": " << targetPhrase << std::endl;
}
TargetPhraseCollectionWithSourcePhrase *rv=PruneTargetCandidates(tCands, costs, sourcePhrases);
if(rv->IsEmpty())
delete rv;
else {
m_rangeCache[i->first.first][i->first.second-1]=rv;
m_tgtColls.push_back(rv);
}
}
// free memory
m_dict->FreeMemory();
}
void PDTAimp::CreateTargetPhrase(TargetPhrase& targetPhrase,
StringTgtCand::Tokens const& factorStrings,
Scores const& transVector,
Scores const& inputVector,
const std::string *alignmentString,
Phrase const* srcPtr) const {
FactorCollection &factorCollection = FactorCollection::Instance();
for(size_t k=0; k<factorStrings.size(); ++k) {
util::TokenIter<util::MultiCharacter, false> word(*factorStrings[k], StaticData::Instance().GetFactorDelimiter());
Word& w=targetPhrase.AddWord();
for(size_t l=0; l<m_output.size(); ++l, ++word) {
w[m_output[l]]= factorCollection.AddFactor(*word);
}
}
if (alignmentString) {
targetPhrase.SetAlignmentInfo(*alignmentString);
}
if (m_numInputScores) {
targetPhrase.GetScoreBreakdown().Assign(m_inputFeature, inputVector);
}
targetPhrase.GetScoreBreakdown().Assign(m_obj, transVector);
targetPhrase.Evaluate(*srcPtr, m_obj->GetFeaturesToApply());
}
TargetPhraseCollectionWithSourcePhrase* PDTAimp::PruneTargetCandidates
(const std::vector<TargetPhrase> & tCands,
std::vector<std::pair<float,size_t> >& costs,
const std::vector<Phrase> &sourcePhrases) const {
// convert into TargetPhraseCollection
UTIL_THROW_IF2(tCands.size() != sourcePhrases.size(),
"Number of target phrases must equal number of source phrases");
TargetPhraseCollectionWithSourcePhrase *rv=new TargetPhraseCollectionWithSourcePhrase;
// set limit to tableLimit or actual size, whatever is smaller
std::vector<std::pair<float,size_t> >::iterator nth =
costs.begin() + ((m_obj->m_tableLimit>0 && // 0 indicates no limit
m_obj->m_tableLimit < costs.size()) ?
m_obj->m_tableLimit : costs.size());
// find the nth phrase according to future cost
NTH_ELEMENT3(costs.begin(),nth ,costs.end());
// add n top phrases to the return list
for(std::vector<std::pair<float,size_t> >::iterator
it = costs.begin(); it != nth; ++it) {
size_t ind = it->second;
TargetPhrase *targetPhrase = new TargetPhrase(tCands[ind]);
const Phrase &sourcePhrase = sourcePhrases[ind];
rv->Add(targetPhrase, sourcePhrase);
}
return rv;
}
}

443
moses/PDTAimp.h
View File

@ -38,23 +38,7 @@ class PDTAimp
friend class PhraseDictionaryTreeAdaptor;
protected:
PDTAimp(PhraseDictionaryTreeAdaptor *p)
: m_dict(0),
m_obj(p),
useCache(1),
totalE(0),
distinctE(0) {
m_numInputScores = 0;
const StaticData &staticData = StaticData::Instance();
m_inputFeature = &InputFeature::Instance();
if (m_inputFeature) {
const PhraseDictionary *firstPt = PhraseDictionary::GetColl()[0];
if (firstPt == m_obj) {
m_numInputScores = m_inputFeature->GetNumScoreComponents();
}
}
}
PDTAimp(PhraseDictionaryTreeAdaptor *p);
public:
std::vector<FactorType> m_input,m_output;
@ -77,172 +61,22 @@ public:
std::vector<size_t> path1Best,pathExplored;
std::vector<double> pathCN;
~PDTAimp() {
CleanUp();
delete m_dict;
if (StaticData::Instance().GetVerboseLevel() >= 2) {
TRACE_ERR("tgt candidates stats: total="<<totalE<<"; distinct="
<<distinctE<<" ("<<distinctE/(0.01*totalE)<<"); duplicates="
<<totalE-distinctE<<" ("<<(totalE-distinctE)/(0.01*totalE)
<<")\n");
TRACE_ERR("\npath statistics\n");
if(path1Best.size()) {
TRACE_ERR("1-best: ");
std::copy(path1Best.begin()+1,path1Best.end(),
std::ostream_iterator<size_t>(std::cerr," \t"));
TRACE_ERR("\n");
}
if(pathCN.size()) {
TRACE_ERR("CN (full): ");
std::transform(pathCN.begin()+1
,pathCN.end()
,std::ostream_iterator<double>(std::cerr," \t")
,Exp);
TRACE_ERR("\n");
}
if(pathExplored.size()) {
TRACE_ERR("CN (explored): ");
std::copy(pathExplored.begin()+1,pathExplored.end(),
std::ostream_iterator<size_t>(std::cerr," \t"));
TRACE_ERR("\n");
}
}
}
~PDTAimp();
void Factors2String(Word const& w,std::string& s) const {
s=w.GetString(m_input,false);
}
void CleanUp() {
assert(m_dict);
m_dict->FreeMemory();
for(size_t i=0; i<m_tgtColls.size(); ++i) delete m_tgtColls[i];
m_tgtColls.clear();
m_cache.clear();
m_rangeCache.clear();
uniqSrcPhr.clear();
}
void CleanUp();
TargetPhraseCollectionWithSourcePhrase const*
GetTargetPhraseCollection(Phrase const &src) const {
assert(m_dict);
if(src.GetSize()==0) return 0;
std::pair<MapSrc2Tgt::iterator,bool> piter;
if(useCache) {
piter=m_cache.insert(std::make_pair(src,static_cast<TargetPhraseCollectionWithSourcePhrase const*>(0)));
if(!piter.second) return piter.first->second;
} else if (m_cache.size()) {
MapSrc2Tgt::const_iterator i=m_cache.find(src);
return (i!=m_cache.end() ? i->second : 0);
}
std::vector<std::string> srcString(src.GetSize());
// convert source Phrase into vector of strings
for(size_t i=0; i<srcString.size(); ++i) {
Factors2String(src.GetWord(i),srcString[i]);
}
// get target phrases in string representation
std::vector<StringTgtCand> cands;
std::vector<std::string> wacands;
m_dict->GetTargetCandidates(srcString,cands,wacands);
if(cands.empty()) {
return 0;
}
//TODO: Multiple models broken here
std::vector<float> weights = StaticData::Instance().GetWeights(m_obj);
std::vector<TargetPhrase> tCands;
tCands.reserve(cands.size());
std::vector<std::pair<float,size_t> > costs;
costs.reserve(cands.size());
std::vector<Phrase> sourcePhrases;
sourcePhrases.reserve(cands.size());
// convert into TargetPhrases
for(size_t i=0; i<cands.size(); ++i) {
TargetPhrase targetPhrase;
StringTgtCand::Tokens const& factorStrings=cands[i].tokens;
Scores const& probVector=cands[i].scores;
std::vector<float> scoreVector(probVector.size());
std::transform(probVector.begin(),probVector.end(),scoreVector.begin(),
TransformScore);
std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),
FloorScore);
//sparse features.
//These are already in log-space
for (size_t j = 0; j < cands[i].fnames.size(); ++j) {
targetPhrase.GetScoreBreakdown().Assign(m_obj, *cands[i].fnames[j], cands[i].fvalues[j]);
}
CreateTargetPhrase(targetPhrase,factorStrings,scoreVector, Scores(0), &wacands[i], &src);
costs.push_back(std::make_pair(-targetPhrase.GetFutureScore(),tCands.size()));
tCands.push_back(targetPhrase);
sourcePhrases.push_back(src);
}
TargetPhraseCollectionWithSourcePhrase *rv;
rv=PruneTargetCandidates(tCands,costs, sourcePhrases);
if(rv->IsEmpty()) {
delete rv;
return 0;
} else {
if(useCache) piter.first->second=rv;
m_tgtColls.push_back(rv);
return rv;
}
}
GetTargetPhraseCollection(Phrase const &src) const;
void Create(const std::vector<FactorType> &input
, const std::vector<FactorType> &output
, const std::string &filePath
, const std::vector<float> &weight
) {
, const std::vector<float> &weight);
// set my members
m_dict=new PhraseDictionaryTree();
m_input=input;
m_output=output;
const StaticData &staticData = StaticData::Instance();
m_dict->NeedAlignmentInfo(staticData.NeedAlignmentInfo());
std::string binFname=filePath+".binphr.idx";
if(!FileExists(binFname.c_str())) {
UTIL_THROW2( "bin ttable does not exist");
//TRACE_ERR( "bin ttable does not exist -> create it\n");
//InputFileStream in(filePath);
//m_dict->Create(in,filePath);
}
VERBOSE(1,"reading bin ttable\n");
// m_dict->Read(filePath);
bool res=m_dict->Read(filePath);
if (!res) {
std::stringstream strme;
strme << "bin ttable was read in a wrong way\n";
UserMessage::Add(strme.str());
exit(1);
}
}
typedef PhraseDictionaryTree::PrefixPtr PPtr;
typedef unsigned short Position;
@ -285,61 +119,13 @@ public:
Scores const& transVector,
Scores const& inputVector,
const std::string *alignmentString,
Phrase const* srcPtr=0) const {
FactorCollection &factorCollection = FactorCollection::Instance();
for(size_t k=0; k<factorStrings.size(); ++k) {
util::TokenIter<util::MultiCharacter, false> word(*factorStrings[k], StaticData::Instance().GetFactorDelimiter());
Word& w=targetPhrase.AddWord();
for(size_t l=0; l<m_output.size(); ++l, ++word) {
w[m_output[l]]= factorCollection.AddFactor(*word);
}
}
if (alignmentString) {
targetPhrase.SetAlignmentInfo(*alignmentString);
}
if (m_numInputScores) {
targetPhrase.GetScoreBreakdown().Assign(m_inputFeature, inputVector);
}
targetPhrase.GetScoreBreakdown().Assign(m_obj, transVector);
targetPhrase.Evaluate(*srcPtr, m_obj->GetFeaturesToApply());
}
Phrase const* srcPtr=0) const;
TargetPhraseCollectionWithSourcePhrase* PruneTargetCandidates
(const std::vector<TargetPhrase> & tCands,
std::vector<std::pair<float,size_t> >& costs,
const std::vector<Phrase> &sourcePhrases) const {
// convert into TargetPhraseCollection
UTIL_THROW_IF2(tCands.size() != sourcePhrases.size(),
"Number of target phrases must equal number of source phrases");
const std::vector<Phrase> &sourcePhrases) const;
TargetPhraseCollectionWithSourcePhrase *rv=new TargetPhraseCollectionWithSourcePhrase;
// set limit to tableLimit or actual size, whatever is smaller
std::vector<std::pair<float,size_t> >::iterator nth =
costs.begin() + ((m_obj->m_tableLimit>0 && // 0 indicates no limit
m_obj->m_tableLimit < costs.size()) ?
m_obj->m_tableLimit : costs.size());
// find the nth phrase according to future cost
NTH_ELEMENT3(costs.begin(),nth ,costs.end());
// add n top phrases to the return list
for(std::vector<std::pair<float,size_t> >::iterator
it = costs.begin(); it != nth; ++it) {
size_t ind = it->second;
TargetPhrase *targetPhrase = new TargetPhrase(tCands[ind]);
const Phrase &sourcePhrase = sourcePhrases[ind];
rv->Add(targetPhrase, sourcePhrase);
}
return rv;
}
// POD for target phrase scores
struct TScores {
@ -350,220 +136,7 @@ public:
TScores() : total(0.0),src(0) {}
};
void CacheSource(ConfusionNet const& src) {
assert(m_dict);
const size_t srcSize=src.GetSize();
std::vector<size_t> exploredPaths(srcSize+1,0);
std::vector<double> exPathsD(srcSize+1,-1.0);
// collect some statistics
std::vector<size_t> cnDepths(srcSize,0);
for(size_t i=0; i<srcSize; ++i) cnDepths[i]=src[i].size();
for(size_t len=1; len<=srcSize; ++len)
for(size_t i=0; i<=srcSize-len; ++i) {
double pd=0.0;
for(size_t k=i; k<i+len; ++k) pd+=log(1.0*cnDepths[k]);
exPathsD[len]=(exPathsD[len]>=0.0 ? addLogScale(pd,exPathsD[len]) : pd);
}
// update global statistics
if(pathCN.size()<=srcSize) pathCN.resize(srcSize+1,-1.0);
for(size_t len=1; len<=srcSize; ++len)
pathCN[len]=pathCN[len]>=0.0 ? addLogScale(pathCN[len],exPathsD[len]) : exPathsD[len];
if(path1Best.size()<=srcSize) path1Best.resize(srcSize+1,0);
for(size_t len=1; len<=srcSize; ++len) path1Best[len]+=srcSize-len+1;
if (StaticData::Instance().GetVerboseLevel() >= 2 && exPathsD.size()) {
TRACE_ERR("path stats for current CN: \nCN (full): ");
std::transform(exPathsD.begin()+1
,exPathsD.end()
,std::ostream_iterator<double>(std::cerr," ")
,Exp);
TRACE_ERR("\n");
}
typedef StringTgtCand::Tokens sPhrase;
typedef std::map<StringTgtCand::Tokens,TScores> E2Costs;
std::map<Range,E2Costs> cov2cand;
std::vector<State> stack;
for(Position i=0 ; i < srcSize ; ++i)
stack.push_back(State(i, i, m_dict->GetRoot(), std::vector<float>(m_numInputScores,0.0)));
std::vector<float> weightTrans = StaticData::Instance().GetWeights(m_obj);
std::vector<float> weightInput = StaticData::Instance().GetWeights(m_inputFeature);
float weightWP = StaticData::Instance().GetWeightWordPenalty();
while(!stack.empty()) {
State curr(stack.back());
stack.pop_back();
UTIL_THROW_IF2(curr.end() >= srcSize, "Error");
const ConfusionNet::Column &currCol=src[curr.end()];
// in a given column, loop over all possibilities
for(size_t colidx=0; colidx<currCol.size(); ++colidx) {
const Word& w=currCol[colidx].first; // w=the i^th possibility in column colidx
std::string s;
Factors2String(w,s);
bool isEpsilon=(s=="" || s==EPSILON);
//assert that we have the right number of link params in this CN option
UTIL_THROW_IF2(currCol[colidx].second.denseScores.size() < m_numInputScores,
"Incorrect number of input scores");
// do not start with epsilon (except at first position)
if(isEpsilon && curr.begin()==curr.end() && curr.begin()>0) continue;
// At a given node in the prefix tree, look to see if w defines an edge to
// another node (Extend). Stay at the same node if w==EPSILON
PPtr nextP = (isEpsilon ? curr.ptr : m_dict->Extend(curr.ptr,s));
if(nextP) { // w is a word that should be considered
Range newRange(curr.begin(),curr.end()+src.GetColumnIncrement(curr.end(),colidx));
//add together the link scores from the current state and the new arc
float inputScoreSum = 0;
std::vector<float> newInputScores(m_numInputScores,0.0);
if (m_numInputScores) {
std::transform(currCol[colidx].second.denseScores.begin(), currCol[colidx].second.denseScores.end(),
curr.GetScores().begin(),
newInputScores.begin(),
std::plus<float>());
//we need to sum up link weights (excluding realWordCount, which isn't in numLinkParams)
//if the sum is too low, then we won't expand this.
//TODO: dodgy! shouldn't we consider weights here? what about zero-weight params?
inputScoreSum = std::accumulate(newInputScores.begin(),newInputScores.begin()+m_numInputScores,0.0);
}
Phrase newSrc(curr.src);
if(!isEpsilon) newSrc.AddWord(w);
if(newRange.second<srcSize && inputScoreSum>LOWEST_SCORE) {
// if there is more room to grow, add a new state onto the queue
// to be explored that represents [begin, curEnd+)
stack.push_back(State(newRange,nextP,newInputScores));
stack.back().src=newSrc;
}
std::vector<StringTgtCand> tcands;
// now, look up the target candidates (aprx. TargetPhraseCollection) for
// the current path through the CN
m_dict->GetTargetCandidates(nextP,tcands);
if(newRange.second>=exploredPaths.size()+newRange.first)
exploredPaths.resize(newRange.second-newRange.first+1,0);
++exploredPaths[newRange.second-newRange.first];
totalE+=tcands.size();
if(tcands.size()) {
E2Costs& e2costs=cov2cand[newRange];
Phrase const* srcPtr=uniqSrcPhr(newSrc);
for(size_t i=0; i<tcands.size(); ++i) {
//put input scores in first - already logged, just drop in directly
std::vector<float> transcores(m_obj->GetNumScoreComponents());
UTIL_THROW_IF2(transcores.size() != weightTrans.size(),
"Incorrect number of translation scores");
//put in phrase table scores, logging as we insert
std::transform(tcands[i].scores.begin()
,tcands[i].scores.end()
,transcores.begin()
,TransformScore);
//tally up
float score=std::inner_product(transcores.begin(), transcores.end(), weightTrans.begin(), 0.0f);
// input feature
score +=std::inner_product(newInputScores.begin(), newInputScores.end(), weightInput.begin(), 0.0f);
//count word penalty
score-=tcands[i].tokens.size() * weightWP;
std::pair<E2Costs::iterator,bool> p=e2costs.insert(std::make_pair(tcands[i].tokens,TScores()));
if(p.second) ++distinctE;
TScores & scores=p.first->second;
if(p.second || scores.total<score) {
scores.total=score;
scores.transScore=transcores;
scores.inputScores=newInputScores;
scores.src=srcPtr;
}
}
}
}
}
} // end while(!stack.empty())
if (StaticData::Instance().GetVerboseLevel() >= 2 && exploredPaths.size()) {
TRACE_ERR("CN (explored): ");
std::copy(exploredPaths.begin()+1,exploredPaths.end(),
std::ostream_iterator<size_t>(std::cerr," "));
TRACE_ERR("\n");
}
if(pathExplored.size()<exploredPaths.size())
pathExplored.resize(exploredPaths.size(),0);
for(size_t len=1; len<=srcSize; ++len)
pathExplored[len]+=exploredPaths[len];
m_rangeCache.resize(src.GetSize(),vTPC(src.GetSize(),0));
for(std::map<Range,E2Costs>::const_iterator i=cov2cand.begin(); i!=cov2cand.end(); ++i) {
assert(i->first.first<m_rangeCache.size());
assert(i->first.second>0);
assert(static_cast<size_t>(i->first.second-1)<m_rangeCache[i->first.first].size());
assert(m_rangeCache[i->first.first][i->first.second-1]==0);
std::vector<TargetPhrase> tCands;
tCands.reserve(i->second.size());
std::vector<std::pair<float,size_t> > costs;
costs.reserve(i->second.size());
std::vector<Phrase> sourcePhrases;
sourcePhrases.reserve(i->second.size());
for(E2Costs::const_iterator j=i->second.begin(); j!=i->second.end(); ++j) {
TScores const & scores=j->second;
TargetPhrase targetPhrase;
CreateTargetPhrase(targetPhrase
, j ->first
, scores.transScore
, scores.inputScores
, NULL
, scores.src);
costs.push_back(std::make_pair(-targetPhrase.GetFutureScore(),tCands.size()));
tCands.push_back(targetPhrase);
sourcePhrases.push_back(*scores.src);
//std::cerr << i->first.first << "-" << i->first.second << ": " << targetPhrase << std::endl;
}
TargetPhraseCollectionWithSourcePhrase *rv=PruneTargetCandidates(tCands, costs, sourcePhrases);
if(rv->IsEmpty())
delete rv;
else {
m_rangeCache[i->first.first][i->first.second-1]=rv;
m_tgtColls.push_back(rv);
}
}
// free memory
m_dict->FreeMemory();
}
void CacheSource(ConfusionNet const& src);
size_t GetNumInputScores() const {
return m_numInputScores;

24
moses/TargetPhrase.cpp
View File

@ -38,24 +38,7 @@ using namespace std;
namespace Moses
{
TargetPhrase::TargetPhrase( std::string out_string)
:Phrase(0)
, m_fullScore(0.0)
, m_futureScore(0.0)
, m_alignTerm(&AlignmentInfoCollection::Instance().GetEmptyAlignmentInfo())
, m_alignNonTerm(&AlignmentInfoCollection::Instance().GetEmptyAlignmentInfo())
, m_lhsTarget(NULL)
, m_ruleSource(NULL)
{
//ACAT
const StaticData &staticData = StaticData::Instance();
CreateFromString(Output, staticData.GetInputFactorOrder(), out_string,
// staticData.GetFactorDelimiter(), // eliminated [UG]
NULL);
}
TargetPhrase::TargetPhrase()
TargetPhrase::TargetPhrase(const PhraseDictionary *pt)
:Phrase()
, m_fullScore(0.0)
, m_futureScore(0.0)
@ -63,10 +46,11 @@ TargetPhrase::TargetPhrase()
, m_alignNonTerm(&AlignmentInfoCollection::Instance().GetEmptyAlignmentInfo())
, m_lhsTarget(NULL)
, m_ruleSource(NULL)
, m_container(pt)
{
}
TargetPhrase::TargetPhrase(const Phrase &phrase)
TargetPhrase::TargetPhrase(const Phrase &phrase, const PhraseDictionary *pt)
: Phrase(phrase)
, m_fullScore(0.0)
, m_futureScore(0.0)
@ -74,6 +58,7 @@ TargetPhrase::TargetPhrase(const Phrase &phrase)
, m_alignNonTerm(&AlignmentInfoCollection::Instance().GetEmptyAlignmentInfo())
, m_lhsTarget(NULL)
, m_ruleSource(NULL)
, m_container(pt)
{
}
@ -84,6 +69,7 @@ TargetPhrase::TargetPhrase(const TargetPhrase &copy)
, m_scoreBreakdown(copy.m_scoreBreakdown)
, m_alignTerm(copy.m_alignTerm)
, m_alignNonTerm(copy.m_alignNonTerm)
, m_container(copy.m_container)
{
if (copy.m_lhsTarget) {
m_lhsTarget = new Word(*copy.m_lhsTarget);

12
moses/TargetPhrase.h
View File

@ -41,6 +41,8 @@ namespace Moses
{
class FeatureFunction;
class InputPath;
class InputPath;
class PhraseDictionary;
/** represents an entry on the target side of a phrase table (scores, translation, alignment)
*/
@ -60,11 +62,12 @@ private:
typedef std::map<std::string, boost::shared_ptr<PhraseProperty> > Properties;
Properties m_properties;
const PhraseDictionary *m_container;
public:
TargetPhrase();
TargetPhrase(const PhraseDictionary *pt = NULL);
TargetPhrase(const TargetPhrase &copy);
explicit TargetPhrase(std::string out_string);
explicit TargetPhrase(const Phrase &targetPhrase);
explicit TargetPhrase(const Phrase &targetPhrase, const PhraseDictionary *pt);
~TargetPhrase();
// 1st evaluate method. Called during loading of phrase table.
@ -132,6 +135,9 @@ public:
return m_ruleSource;
}
const PhraseDictionary *GetContainer() const
{ return m_container; }
// To be set by the FF that needs it, by default the rule source = NULL
// make a copy of the source side of the rule
void SetRuleSource(const Phrase &ruleSource) const;

4
moses/TranslationModel/BilingualDynSuffixArray.cpp
View File

@ -377,9 +377,9 @@ TrgPhraseFromSntIdx(const PhrasePair& phrasepair) const
TargetPhrase*
BilingualDynSuffixArray::
GetMosesFactorIDs(const SAPhrase& phrase, const Phrase& sourcePhrase) const
GetMosesFactorIDs(const SAPhrase& phrase, const Phrase& sourcePhrase, const PhraseDictionary *pt) const
{
TargetPhrase* targetPhrase = new TargetPhrase();
TargetPhrase* targetPhrase = new TargetPhrase(pt);
for(size_t i=0; i < phrase.words.size(); ++i) { // look up trg words
Word& word = m_trgVocab->GetWord( phrase.words[i]);
UTIL_THROW_IF2(word == m_trgVocab->GetkOOVWord(),

2
moses/TranslationModel/BilingualDynSuffixArray.h
View File

@ -128,7 +128,7 @@ public:
GatherCands(Phrase const& src, map<SAPhrase, vector<float> >& pstats) const;
TargetPhrase*
GetMosesFactorIDs(const SAPhrase&, const Phrase& sourcePhrase) const;
GetMosesFactorIDs(const SAPhrase&, const Phrase& sourcePhrase, const PhraseDictionary *pt) const;
private:

2
moses/TranslationModel/CYKPlusParser/ChartRuleLookupManagerSkeleton.cpp
View File

@ -77,7 +77,7 @@ TargetPhrase *ChartRuleLookupManagerSkeleton::CreateTargetPhrase(const Word &sou
string str = sourceWord.GetFactor(0)->GetString().as_string();
str = "ChartManagerSkeleton:" + str;
TargetPhrase *tp = new TargetPhrase();
TargetPhrase *tp = new TargetPhrase(&m_skeletonPT);
Word &word = tp->AddWord();
word.CreateFromString(Output, m_skeletonPT.GetOutput(), str, false);

2
moses/TranslationModel/CompactPT/PhraseDecoder.cpp
View File

@ -418,7 +418,7 @@ TargetPhraseVectorPtr PhraseDecoder::DecodeCollection(
}
if(eval) {
targetPhrase->Evaluate(sourcePhrase);
targetPhrase->Evaluate(sourcePhrase, m_phraseDictionary.GetFeaturesToApply());
}
if(m_coding == PREnc) {

2
moses/TranslationModel/CompactPT/PhraseDictionaryCompact.cpp
View File

@ -57,6 +57,8 @@ void PhraseDictionaryCompact::Load()
{
const StaticData &staticData = StaticData::Instance();
SetFeaturesToApply();
m_weight = staticData.GetWeights(this);
std::string tFilePath = m_filePath;

2
moses/TranslationModel/PhraseDictionary.cpp
View File

@ -24,6 +24,7 @@ Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
#include "moses/InputType.h"
#include "moses/TranslationOption.h"
#include "moses/UserMessage.h"
#include "moses/DecodeStep.h"
#include "moses/DecodeGraph.h"
#include "moses/InputPath.h"
#include "util/exception.hh"
@ -48,6 +49,7 @@ PhraseDictionary::PhraseDictionary(const std::string &line)
,m_tableLimit(20) // default
,m_maxCacheSize(DEFAULT_MAX_TRANS_OPT_CACHE_SIZE)
{
m_id = s_staticColl.size();
s_staticColl.push_back(this);
}

5
moses/TranslationModel/PhraseDictionary.h
View File

@ -87,6 +87,10 @@ public:
return m_tableLimit;
}
//! continguous id for each pt, starting from 0
size_t GetId() const
{ return m_id; }
virtual
void
Release(TargetPhraseCollection const* tpc) const;
@ -167,6 +171,7 @@ protected:
protected:
CacheColl &GetCache() const;
size_t m_id;
};

2
moses/TranslationModel/PhraseDictionaryDynSuffixArray.cpp
View File

@ -58,7 +58,7 @@ GetTargetPhraseCollectionLEGACY(const Phrase& src) const
TargetPhraseCollection *ret = new TargetPhraseCollection();
BOOST_FOREACH(pstat_entry & e, pstats) {
TargetPhrase* tp = m_biSA->GetMosesFactorIDs(e.first, src);
TargetPhrase* tp = m_biSA->GetMosesFactorIDs(e.first, src, this);
tp->GetScoreBreakdown().Assign(this,e.second);
tp->Evaluate(src);
ret->Add(tp);

2
moses/TranslationModel/PhraseDictionaryTransliteration.cpp
View File

@ -124,7 +124,7 @@ std::vector<TargetPhrase*> PhraseDictionaryTransliteration::CreateTargetPhrases(
Tokenize(toks, line, "\t");
UTIL_THROW_IF2(toks.size() != 2, "Error in transliteration output file. Expecting word\tscore");
TargetPhrase *tp = new TargetPhrase();
TargetPhrase *tp = new TargetPhrase(this);
Word &word = tp->AddWord();
word.CreateFromString(Output, m_output, toks[0], false);

2
moses/TranslationModel/ProbingPT/ProbingPT.cpp
View File

@ -153,7 +153,7 @@ TargetPhrase *ProbingPT::CreateTargetPhrase(const Phrase &sourcePhrase, const ta
const std::vector<unsigned int> &probingPhrase = probingTargetPhrase.target_phrase;
size_t size = probingPhrase.size();
TargetPhrase *tp = new TargetPhrase();
TargetPhrase *tp = new TargetPhrase(this);
// words
for (size_t i = 0; i < size; ++i) {

2
moses/TranslationModel/RuleTable/LoaderCompact.cpp
View File

@ -222,7 +222,7 @@ bool RuleTableLoaderCompact::LoadRuleSection(
// The remaining columns are currently ignored.
// Create and score target phrase.
TargetPhrase *targetPhrase = new TargetPhrase(targetPhrasePhrase);
TargetPhrase *targetPhrase = new TargetPhrase(targetPhrasePhrase, &ruleTable);
targetPhrase->SetAlignNonTerm(alignNonTerm);
targetPhrase->SetTargetLHS(targetLhs);

2
moses/TranslationModel/RuleTable/LoaderStandard.cpp
View File

@ -222,7 +222,7 @@ bool RuleTableLoaderStandard::Load(FormatType format
Word *targetLHS;
// create target phrase obj
TargetPhrase *targetPhrase = new TargetPhrase();
TargetPhrase *targetPhrase = new TargetPhrase(&ruleTable);
// targetPhrase->CreateFromString(Output, output, targetPhraseString, factorDelimiter, &targetLHS);
targetPhrase->CreateFromString(Output, output, targetPhraseString, &targetLHS);
// source

2
moses/TranslationModel/RuleTable/PhraseDictionaryFuzzyMatch.cpp
View File

@ -270,7 +270,7 @@ void PhraseDictionaryFuzzyMatch::InitializeForInput(InputType const& inputSenten
sourcePhrase.CreateFromString(Input, m_input, sourcePhraseString, &sourceLHS);
// create target phrase obj
TargetPhrase *targetPhrase = new TargetPhrase();
TargetPhrase *targetPhrase = new TargetPhrase(this);
// targetPhrase->CreateFromString(Output, m_output, targetPhraseString, factorDelimiter, &targetLHS);
targetPhrase->CreateFromString(Output, m_output, targetPhraseString, &targetLHS);

2
moses/TranslationModel/SkeletonPT.cpp
View File

@ -53,7 +53,7 @@ TargetPhrase *SkeletonPT::CreateTargetPhrase(const Phrase &sourcePhrase) const
string str = sourcePhrase.GetWord(0).GetFactor(0)->GetString().as_string();
str = "SkeletonPT:" + str;
TargetPhrase *tp = new TargetPhrase();
TargetPhrase *tp = new TargetPhrase(this);
Word &word = tp->AddWord();
word.CreateFromString(Output, m_output, str, false);

2
moses/TranslationModel/UG/mmsapt.cpp
View File

@ -466,7 +466,7 @@ namespace Moses
BOOST_FOREACH(sptr<pscorer> const& ff, m_active_ff_common)
(*ff)(*dynbt, pool, &fvals);
}
TargetPhrase* tp = new TargetPhrase();
TargetPhrase* tp = new TargetPhrase(this);
Token const* x = fix ? fix->start2 : dyn->start2;
uint32_t len = fix ? fix->len2 : dyn->len2;
for (uint32_t k = 0; k < len; ++k, x = x->next())

2
moses/TranslationOption.cpp
View File

@ -32,7 +32,7 @@ namespace Moses
{
TranslationOption::TranslationOption()
:m_targetPhrase()
:m_targetPhrase(NULL)
,m_inputPath(NULL)
,m_sourceWordsRange(NOT_FOUND, NOT_FOUND)
{

2
moses/TranslationOptionCollection.cpp
View File

@ -231,7 +231,7 @@ void TranslationOptionCollection::ProcessOneUnknownWord(const InputPath &inputPa
// modify the starting bitmap
}
TargetPhrase targetPhrase;
TargetPhrase targetPhrase(NULL);
if (!(staticData.GetDropUnknown() || isEpsilon) || isDigit) {
// add to dictionary

2
moses/TreeInput.cpp
View File

@ -173,7 +173,7 @@ bool TreeInput::ProcessAndStripXMLTags(string &line, std::vector<XMLParseOutput>
//TRACE_ERR("number of translations: " << altTexts.size() << endl);
for (size_t i=0; i<altTexts.size(); ++i) {
// set target phrase
TargetPhrase targetPhrase;
TargetPhrase targetPhrase(NULL);
// targetPhrase.CreateFromString(Output, outputFactorOrder,altTexts[i],factorDelimiter, NULL);
targetPhrase.CreateFromString(Output, outputFactorOrder,altTexts[i], NULL);

2
moses/XmlOption.cpp
View File

@ -361,7 +361,7 @@ bool ProcessAndStripXMLTags(string &line, vector<XmlOption*> &res, ReorderingCon
float scoreValue = FloorScore(TransformScore(probValue));
WordsRange range(startPos + offset,endPos-1 + offset); // span covered by phrase
TargetPhrase targetPhrase;
TargetPhrase targetPhrase(NULL);
// targetPhrase.CreateFromString(Output, outputFactorOrder,altTexts[i],factorDelimiter, NULL);
targetPhrase.CreateFromString(Output, outputFactorOrder,altTexts[i], NULL);

1
phrase-extract/extract-ghkm/PhraseOrientation.cpp
View File

@ -20,6 +20,7 @@
#include "PhraseOrientation.h"
#include <iostream>
#include <sstream>
#include <limits>
#include <boost/assign/list_of.hpp>