Hi, simple answer: all phrase pairs that are consistent with the word alignment are collected during training.
For more detail, read the section "Methods for Learning Phrase Translations" on this page: http://www.statmt.org/moses/?n=Moses.Background -phi On Mon, Mar 18, 2013 at 3:43 PM, Per Tunedal <[email protected]> wrote: > Hi, > I'm pondering about how to best prepare my corpus. That's why I'm > interested in how Moses picks the phrases when training for a phrase > model. Any short explanation in plain wording? Any reading? > Yours, > Per Tunedal > _______________________________________________ > 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
