moses-smt/mosesdecoder
1
1
Fork 0
mirror of https://github.com/moses-smt/mosesdecoder.git synced 2024-12-25 04:43:03 +03:00
Code Issues Packages Projects Releases Wiki Activity

Proper spacing

Browse Source
This commit is contained in:
alvations 2020-01-06 11:43:31 +08:00 committed by GitHub
parent f46ee7c5ac
commit d03df21e88
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 50 additions and 50 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

100
scripts/ems/support/split-sentences.perl
View File

@ -27,10 +27,10 @@ my $LIST_ITEM = 0;
while (@ARGV) { while (@ARGV) {
$_ = shift; $_ = shift;
/^-l$/ && ($language = shift, next); /^-l$/ && ($language = shift, next);
/^-p$/ && ($prefixfile = shift, next); /^-p$/ && ($prefixfile = shift, next);
/^-q$/ && ($QUIET = 1, next); /^-q$/ && ($QUIET = 1, next);
/^-h$/ && ($HELP = 1, next); /^-h$/ && ($HELP = 1, next);
/^-i$/ && ($LIST_ITEM = 1, next); /^-i$/ && ($LIST_ITEM = 1, next);
/^-b$/ && ($|++, next); # no output buffering /^-b$/ && ($|++, next); # no output buffering
} }
@ -39,7 +39,7 @@ if ($HELP) {
print "-q: quiet mode\n"; print "-q: quiet mode\n";
print "-b: no output buffering (for use in bidirectional pipes)\n"; print "-b: no output buffering (for use in bidirectional pipes)\n";
print "-p: use a custom prefix file, overriding the installed one\n"; print "-p: use a custom prefix file, overriding the installed one\n";
print "-i: avoid splitting on list items (e.g. 1. This is the first)\n"; print "-i: avoid splitting on list items (e.g. 1. This is the first)\n";
exit; exit;
} }
if (!$QUIET) { if (!$QUIET) {
@ -53,17 +53,17 @@ if ($language eq "yue" || $language eq "zh" || $language eq "ja") {
} }
if ($prefixfile ne "") { if ($prefixfile ne "") {
print STDERR "Loading non-breaking prefixes from $prefixfile\n"; print STDERR "Loading non-breaking prefixes from $prefixfile\n";
} else { } else {
$prefixfile = "$mydir/nonbreaking_prefix.$language"; $prefixfile = "$mydir/nonbreaking_prefix.$language";
# Default to English, if we don't have a language-specific prefix file. # Default to English, if we don't have a language-specific prefix file.
if (!(-e $prefixfile)) { if (!(-e $prefixfile)) {
$prefixfile = "$mydir/nonbreaking_prefix.en"; $prefixfile = "$mydir/nonbreaking_prefix.en";
print STDERR "WARNING: No known abbreviations for language '$language', attempting fall-back to English version...\n"; print STDERR "WARNING: No known abbreviations for language '$language', attempting fall-back to English version...\n";
die ("ERROR: No abbreviations files found in $mydir\n") unless (-e $prefixfile); die ("ERROR: No abbreviations files found in $mydir\n") unless (-e $prefixfile);
} }
} }
if (-e "$prefixfile") { if (-e "$prefixfile") {
@ -122,21 +122,20 @@ sub preprocess {
##### Add sentence breaks as needed ##### ##### Add sentence breaks as needed #####
# Sentences can start with upper-case, numnbers, or Indic characters # Sentences can start with upper-case, numnbers, or Indic characters
my $sentence_start = "\\p{IsUpper}0-9"; my $sentence_start = "\\p{IsUpper}0-9";
$sentence_start .= "\\p{Block: Devanagari}\\p{Block: Devanagari_Extended}" if ($language eq "hi" || $language eq "mr"); $sentence_start .= "\\p{Block: Devanagari}\\p{Block: Devanagari_Extended}" if ($language eq "hi" || $language eq "mr");
$sentence_start .= "\\p{Block: Gujarati}" if $language eq "gu"; $sentence_start .= "\\p{Block: Gujarati}" if $language eq "gu";
$sentence_start .= "\\p{Block: Bengali}" if ($language eq "as" || $language eq "bn" || $language eq "mni"); $sentence_start .= "\\p{Block: Bengali}" if ($language eq "as" || $language eq "bn" || $language eq "mni");
$sentence_start .= "\\p{Block: Kannada}" if $language eq "kn"; $sentence_start .= "\\p{Block: Kannada}" if $language eq "kn";
$sentence_start .= "\\p{Block: Malayalam}" if $language eq "ml"; $sentence_start .= "\\p{Block: Malayalam}" if $language eq "ml";
$sentence_start .= "\\p{Block: Oriya}" if $language eq "or"; $sentence_start .= "\\p{Block: Oriya}" if $language eq "or";
$sentence_start .= "\\p{Block: Gurmukhi}" if $language eq "pa"; $sentence_start .= "\\p{Block: Gurmukhi}" if $language eq "pa";
$sentence_start .= "\\p{Block: Tamil}" if $language eq "ta"; $sentence_start .= "\\p{Block: Tamil}" if $language eq "ta";
$sentence_start .= "\\p{Block: Telugu}" if $language eq "te"; $sentence_start .= "\\p{Block: Telugu}" if $language eq "te";
$sentence_start .= "\\p{Block: Hangul}\\p{Block: Hangul_Compatibility_Jamo}\\p{Block: Hangul_Jamo}\\p{Block: Hangul_Jamo_Extended_A}\\p{Block: Hangul_Jamo_Extended_B}" if $language eq "ko"; $sentence_start .= "\\p{Block: Hangul}\\p{Block: Hangul_Compatibility_Jamo}\\p{Block: Hangul_Jamo}\\p{Block: Hangul_Jamo_Extended_A}\\p{Block: Hangul_Jamo_Extended_B}" if $language eq "ko";
# we include danda and double danda (U+0964 and U+0965) as sentence split characters
# we include danda and double danda (U+0964 and U+0965) as sentence split characters
# Non-period end of sentence markers (?!) followed by sentence starters. # Non-period end of sentence markers (?!) followed by sentence starters.
$text =~ s/([?!\x{0964}\x{0965}]) +([\'\"\(\[\¿\¡\p{IsPi}]*[$sentence_start])/$1\n$2/g; $text =~ s/([?!\x{0964}\x{0965}]) +([\'\"\(\[\¿\¡\p{IsPi}]*[$sentence_start])/$1\n$2/g;
@ -153,15 +152,15 @@ sub preprocess {
# and are followed by a sentence starter punctuation and upper case. # and are followed by a sentence starter punctuation and upper case.
$text =~ s/([?!\.\x{0964}\x{0965}]) +([\x{300d}\x{300f}\'\"\(\[\¿\¡\p{IsPi}]+[\ ]*[$sentence_start])/$1\n$2/g; $text =~ s/([?!\.\x{0964}\x{0965}]) +([\x{300d}\x{300f}\'\"\(\[\¿\¡\p{IsPi}]+[\ ]*[$sentence_start])/$1\n$2/g;
#NOTE: Korean no longer handled here. #NOTE: Korean no longer handled here, cos Korean has spaces.
if ($is_cjk == 1) { if ($is_cjk == 1) {
# Chinese uses unusual end-of-sentence markers. These are NOT # Chinese uses unusual end-of-sentence markers. These are NOT
# followed by whitespace. Nor is there any idea of capitalization. # followed by whitespace. Nor is there any idea of capitalization.
# There does not appear to be any unicode category for full-stops # There does not appear to be any unicode category for full-stops
# in general, so list them here. U+3002 U+FF0E U+FF1F U+FF01 # in general, so list them here. U+3002 U+FF0E U+FF1F U+FF01
#$text =~ s/([。．？！♪])/$1\n/g; #$text =~ s/([。．？！♪])/$1\n/g;
$text =~ s/([\x{3002}\x{ff0e}\x{FF1F}\x{FF01}]+\s*["\x{201d}\x{201e}\x{300d}\x{300f}]?\s*)/$1\n/g; $text =~ s/([\x{3002}\x{ff0e}\x{FF1F}\x{FF01}]+\s*["\x{201d}\x{201e}\x{300d}\x{300f}]?\s*)/$1\n/g;
# A normal full-stop or other Western sentence enders followed # A normal full-stop or other Western sentence enders followed
# by an ideograph is an end-of-sentence, always. # by an ideograph is an end-of-sentence, always.
@ -175,33 +174,34 @@ sub preprocess {
# spaces here, so that later processing stages can tokenize readily. # spaces here, so that later processing stages can tokenize readily.
# Note that this handles mixed latinate+CJK. # Note that this handles mixed latinate+CJK.
# TODO: perhaps also CJKExtA CJKExtB etc ??? CJK_Radicals_Sup ? # TODO: perhaps also CJKExtA CJKExtB etc ??? CJK_Radicals_Sup ?
# bhaddow - Comment this out since it adds white-space between Chinese characters. This is not
# what we want from sentence-splitter! # bhaddow - Comment this out since it adds white-space between Chinese characters. This is not
# what we want from sentence-splitter!
#$text =~ s/(\p{Punctuation}) *(\p{CJK})/ $1 $2/g; #$text =~ s/(\p{Punctuation}) *(\p{CJK})/ $1 $2/g;
#$text =~ s/(\p{CJK}) *(\p{Punctuation})/$1 $2 /g; #$text =~ s/(\p{CJK}) *(\p{Punctuation})/$1 $2 /g;
#$text =~ s/([\p{CJK}\p{CJKSymbols}])/ $1 /g; #$text =~ s/([\p{CJK}\p{CJKSymbols}])/ $1 /g;
#$text =~ s/ +/ /g; #$text =~ s/ +/ /g;
} }
# Urdu support # Urdu support
# https://en.wikipedia.org/wiki/Urdu_alphabet#Encoding_Urdu_in_Unicode # https://en.wikipedia.org/wiki/Urdu_alphabet#Encoding_Urdu_in_Unicode
if ($language eq 'ur') { if ($language eq 'ur') {
$text =~ s{ $text =~ s{
( (?: [\.\?!\x{06d4}] | \.\.+ ) ( (?: [\.\?!\x{06d4}] | \.\.+ )
[\'\"\x{201e}\x{bb}\(\[\¿\¡\p{IsPf}]* [\'\"\x{201e}\x{bb}\(\[\¿\¡\p{IsPf}]*
) )
\s+ \s+
( [\'\"\x{201e}\x{bb}\(\[\¿\¡\p{IsPi}]* ( [\'\"\x{201e}\x{bb}\(\[\¿\¡\p{IsPi}]*
[\x{0600}-\x{06ff}] [\x{0600}-\x{06ff}]
) )
}{$1\n$2}gx; }{$1\n$2}gx;
} }
# Special punctuation cases are covered. Check all remaining periods. # Special punctuation cases are covered. Check all remaining periods.
my $word; my $word;
my $i; my $i;
my @words = split(/\h/,$text); my @words = split(/\h/,$text);
#print "NOW $text\n"; #print "NOW $text\n";
$text = ""; $text = "";
for ($i=0;$i<(scalar(@words)-1);$i++) { for ($i=0;$i<(scalar(@words)-1);$i++) {
#print "Checking $words[$i] $words[$i+1]\n"; #print "Checking $words[$i] $words[$i+1]\n";
@ -211,19 +211,19 @@ sub preprocess {
my $starting_punct = $2; my $starting_punct = $2;
if ($prefix && $NONBREAKING_PREFIX{$prefix} && $NONBREAKING_PREFIX{$prefix} == 1 && !$starting_punct) { if ($prefix && $NONBREAKING_PREFIX{$prefix} && $NONBREAKING_PREFIX{$prefix} == 1 && !$starting_punct) {
# Not breaking; # Not breaking;
# print "NBP1 $words[$i] $words[$i+1]\n"; ## print "NBP1 $words[$i] $words[$i+1]\n";
} elsif ($words[$i] =~ /(\.)[\p{IsUpper}\-]+(\.+)$/) { } elsif ($words[$i] =~ /(\.)[\p{IsUpper}\-]+(\.+)$/) {
# Not breaking - upper case acronym # Not breaking - upper case acronym
# print "NBP2 $words[$i] $words[$i+1]\n"; #print "NBP2 $words[$i] $words[$i+1]\n";
} elsif ($LIST_ITEM } elsif ($LIST_ITEM
&& ($i == 0 || substr($words[$i-1], -1) eq "\n") && ($i == 0 || substr($words[$i-1], -1) eq "\n")
&& $words[$i] =~ /^\(?(([0-9]+)|([ivx]+)|([A-Za-z]))\)?\.$/) { && $words[$i] =~ /^\(?(([0-9]+)|([ivx]+)|([A-Za-z]))\)?\.$/) {
#Maybe list item - non breaking # Maybe list item - non breaking
# print "NBP3 $words[$i] $words[$i+1]\n"; #print "NBP3 $words[$i] $words[$i+1]\n";
} elsif($words[$i+1] =~ /^([ ]*[\'\"\(\[\¿\¡\p{IsPi}]*[ ]*[0-9$sentence_start])/) { } elsif($words[$i+1] =~ /^([ ]*[\'\"\(\[\¿\¡\p{IsPi}]*[ ]*[0-9$sentence_start])/) {
# The next word has a bunch of initial quotes, maybe a # The next word has a bunch of initial quotes, maybe a
# space, then either upper case or a number # space, then either upper case or a number
# print "MAYBE $words[$i] $words[$i+1]\n"; #print "MAYBE $words[$i] $words[$i+1]\n";
$words[$i] = $words[$i]."\n" unless ($prefix && $NONBREAKING_PREFIX{$prefix} && $NONBREAKING_PREFIX{$prefix} == 2 && !$starting_punct && ($words[$i+1] =~ /^[0-9]+/)); $words[$i] = $words[$i]."\n" unless ($prefix && $NONBREAKING_PREFIX{$prefix} && $NONBREAKING_PREFIX{$prefix} == 2 && !$starting_punct && ($words[$i+1] =~ /^[0-9]+/));
# We always add a return for these, unless we have a # We always add a return for these, unless we have a
# numeric non-breaker and a number start. # numeric non-breaker and a number start.