ah, looking at the code, max-unknowns allows a certain number of OOV, not edit distance
On 20 January 2015 at 14:44, Jesús González Rubio <[email protected]> wrote: > Well, at least 'max_unknowns' seems to not work properly. I've tried > setting it to quite large values (and also -1) and I always obtain the same > result as using zero. E.g. using max_unknowns=-1: > > Translating: what will they do ? CSSD lacks knowledge of both Voldemort > and candy bars in Prague > Line 0: Initialize search took 0.001 seconds total > Total translation options: 346 > Total translation options pruned: 0 > Line 0: Collecting options took 0.001 seconds at moses/Manager.cpp:117 > Stack sizes: 1, 12, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 6, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 0, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Stack sizes: 1, 12, 24, 22, 12, 5, 3, 4, 3, 2, 0, 0, 0, 0, 0, 0, 0 > Line 0: Search took 0.074 seconds > total hypotheses considered = 8606 > number popped from cube = 0 > number not built = 0 > number discarded early = 0 > number discarded = 8516 > number recombined = 2 > number pruned = 0 > time to collect opts 0.001 (0%) > create hyps 0.009 (11%) > estimate score 0.007 (9%) > calc lm 0.000 (0%) > other hyp score 0.018 (24%) > set up cubes 0.000 (0%) > manage cubes 0.000 (0%) > manage stacks 0.008 (11%) > other 0.032 (41%) > total source words = 16 > words deleted = 0 () > words inserted = 0 () > NO BEST TRANSLATION > > 2015-01-20 14:27 GMT+00:00 Hieu Hoang <[email protected]>: > >> max-unknowns = like the edit distance. The maximum number of non-matching >> words >> >> negate = instead of allowing only translations that match the reference, >> allow only translations that do NOT match >> >> soft = don't throw away non-matching translations, just penalise it. >> >> I was basically messing around, I didn't check the results very >> carefully. there may be bugs. >> >> >> On 20 January 2015 at 14:10, Jesús González Rubio < >> [email protected]> wrote: >> >>> Hi all, >>> >>> I have some questions about the constrained decoding feature implemented >>> in moses, (ConstrainedDecoding). >>> >>> Which is the meaning of the 'max-unknowns' parameter? >>> >>> I understand 'max-unknowns' as something like the maximum edit distance >>> allowed between the final translation and the reference, i.e the maximum >>> number of words in the final translation that are allowed to be different >>> to the reference. Is this interpretation correct? >>> >>> Also, what is the interpretation of the 'negate' and 'soft' parameters? >>> >>> Thanks in advance. >>> >>> Cheers. >>> -- >>> Jesús >>> >>> _______________________________________________ >>> Moses-support mailing list >>> [email protected] >>> http://mailman.mit.edu/mailman/listinfo/moses-support >>> >>> >> >> >> -- >> Hieu Hoang >> Research Associate >> University of Edinburgh >> http://www.hoang.co.uk/hieu >> >> > > > -- > Jesús > > _______________________________________________ > Moses-support mailing list > [email protected] > http://mailman.mit.edu/mailman/listinfo/moses-support > > -- Hieu Hoang Research Associate University of Edinburgh http://www.hoang.co.uk/hieu
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
