Dear moses team: I have been (trying) to play with moses specifying multiple decoding paths. However, I have encountered a problem when running the mert script.
In my moses.ini, I specify (as described in an email to this list by Philipp): [input-factors] [mapping] 0 T 0 1 T 1 [ttable-file] 0 0 5 /path/to/ttable1 0 0 5 /path/to/ttable2 [blablabla] [weight-t] 0.2 0.2 0.2 ... 0.2 (10 times) [more blablabla] However, my issue is that, when the mert-moses.pl script runs filter-model-given-input.pl it puts both filtered ttables in exactly the same palce, i.e. it overwrites the first one when it outputs the second one. I have been looking into the code in filter-moses-given-input.pl, and I am wondering if I should be stating in moses.ini: [ttable-file] 0 0 5 /path/to/ttable1 0 1 5 /path/to/ttable2 Is this so? Would this configuration produce the model I expect, i.e. two completely different decoding paths, each with its own set of weights and each table containing a different set of segments? On the other hand, is there some place where I can lookup the moses.ini syntax? Thank you very much in advance, best regards, Germán Sanchis ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program. _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
