We just published an update to corpus filtergraph. The new release, v
3.1.41, fixes a few bugs, enhances ability to run plugins with same name in
different folders independently, and includes total of 35 plugins and 6
example graphs:

   5 additional language filter plugins
   1 new aligner plugin -- wrapper for Champollion Aligner


   $./lib/sa.champollion     Shows how to use Champollion phrase/
                             sentence aligner on sample Chinese data
                             (included). The aligner-champollion.py
                             kernel is complete. Supporting language-
                             specific plugins are in progress and/or
                             need help (volunteers anyone).



   $./lib/extract.dictionary Shows how to extract matched dictionary
                             using filters with the same name in 
                             different folders



Thanks.
Tom


On Mon, 19 Jul 2010 02:08:22 -0700,
<[email protected]>
wrote:
> I invite the Moses community to review the new release of Corpus
> Filtergraph version 3.1.131. The release includes significant updates
> including 29 plugins, four graph examples, and example data to run
scripts.
> This version obsoletes previous plugins.
> 
> Plugins:
> 
>  1 reader class plugin to read input files
> 
>  1 writer class plugin to write output files
> 
> 22 language-agnostic filter class plugins
> 
>  2 language-specific filter class plugins (for Japanese)
>  3 aligner classes plugins to process source and target
>    files in parallel
> 
> Graphs:
>  $./lib/corpus.lm          Shows how to process monolingual
>                            language model data
> 
>  $./lib/corpus.europarl5   Shows how to clean and filter parallel
>                            corpus from sentence-aligned data
> 
>  $./lib/build.europarl5    Shows how to build Moses training data.
>                            Sets include 10 files: 2 parallel 
>                            corpus files, 2 mert tuning files, 2
>                            eval text files and 4 mteval .xml files
>                            for bi-directional testing
>                            (note: requires Moses Decoder toolkit
>                            installed on the host machine because 
>                            one plugin uses tokenizer.perl. Does 
>                            not run on MS Windows)
> 
> https://sourceforge.net/projects/corpfiltergraph/
> 
> Best regards,
> Tom
> _______________________________________________
> Moses-support mailing list
> [email protected]
> http://mailman.mit.edu/mailman/listinfo/moses-support
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to