Hi Rubén,
Are you talking about MMM's Windows TMX extractor? If so, check your TMX file(s) to make sure they're not corrupted in some way, like not having equal TUV's within each TU. "ready-to-use" is always a subjective measure. TMX extractors have various degrees of support for the TMX format. Plus, the extractor can't improve on data that's inappropriate for SMT. TM's that come from translation memory systems contain a lot of data that's not relevant to SMT training. Tom On Thu, 14 Jul 2011 16:39:41 +0200, Rubén de la Fuente wrote: Dear all, I am playing around with Moses for mere mortals (0.991). I have a very large EN>ES TM. When I try to extract the corpus with the Windows add-in, it turns out EN file has 255 559 lines as opposed to 262 858 in the ES one. This will cause problems in the training phase. Any idea how to tackle it? I thought the Windows add-in would provide ready-to-use corpora... Thanks in advance for your help. Cheers, -- Rubén de la Fuente EN/FR>ES translator [email protected] [1] +34 686 33 59 97 skype: rudelafuente twitter: rubendelafuente [2] [3]http://es.linkedin.com/in/rubendelafuente [4] www.wordbonds.es [5] Wordbonds in Facebook [6] Links: ------ [1] mailto:[email protected] [2] http://twitter.com/rubendelafuente [3] http://twitter.com/rubendelafuente [4] http://es.linkedin.com/in/rubendelafuente [5] http://www.wordbonds.es [6] http://www.facebook.com/#!/pages/wordbonds/127510957570
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
