mirror of
https://github.com/moses-smt/mosesdecoder.git
synced 2024-12-28 14:32:38 +03:00
63 lines
2.0 KiB
Plaintext
63 lines
2.0 KiB
Plaintext
README - 16 Jan 2011b
|
|
Author: Jason Riesa <jason.riesa@gmail.com>
|
|
|
|
Picaro [v1.0]: A simple command-line alignment visualization tool.
|
|
Visualize alignments in grid-format.
|
|
|
|
This brief README is organized as follows:
|
|
I. REQUIREMENTS
|
|
II. USAGE
|
|
III. INPUT FORMAT
|
|
IV. EXAMPLE USAGE
|
|
V. NOTES
|
|
|
|
I. REQUIREMENTS
|
|
===============
|
|
Python v2.5 or higher is required.
|
|
|
|
II. USAGE
|
|
=========
|
|
Picaro takes as input 3 mandatory arguments and up to 2 optional arguments:
|
|
Mandatory arguments:
|
|
1. -a1 <alignment1> where alignment1 is a path to an alignment file
|
|
2. -e <e> where e is a path to a file of English sentences
|
|
3. -f <f> where f is a path to a file of French sentences
|
|
Optional arguments:
|
|
1. -a2 <a2> path to alignment2 file in f-e format
|
|
2. -maxlen <len> for each sentence pair, render only when each
|
|
sentence has length in words <= len
|
|
|
|
For historical reasons we use the labels e, f, English, and French,
|
|
but any language pair will do.
|
|
|
|
III. INPUT FORMAT
|
|
=================
|
|
- Files e and f must be sentence-aligned
|
|
- Alignment files must be in f-e format
|
|
See included sample files in zh/ and es/.
|
|
|
|
IV. EXAMPLE USAGE
|
|
=================
|
|
WITH A SINGLE ALIGNMENT:
|
|
$ picaro.py -e zh/sample.e -f zh/sample.f -a1 zh/sample.aln
|
|
|
|
COMPARING TWO ALIGNMENTS:
|
|
$ picaro.py -e zh/sample.e -f zh/sample.f -a1 zh/alternate.aln -a2 zh/sample.aln
|
|
|
|
When visualizing two alignments at once, refer to the following color scheme:
|
|
Green blocks: alignments a1 and a2 agree
|
|
Blue blocks: alignment a1 only
|
|
Gold blocks: alignment a2 only
|
|
|
|
V. NOTES
|
|
========
|
|
RIGHT-TO-LEFT TEXT:
|
|
If you are using right-to-left text, e.g. Arabic, transliterate your text first.
|
|
Terminals generally render unexpectedly with mixed left-to-right and right-to-left text.
|
|
For Arabic, in particular, we use the Buckwalter translitation scheme [1] when using this tool.
|
|
The following Perl module implements Buckwalter transliteration:
|
|
http://search.cpan.org/~smrz/Encode-Arabic-1.8/lib/Encode/Arabic.pm
|
|
|
|
[1] http://www.ldc.upenn.edu/myl/morph/buckwalter.html
|
|
|