.Bold { font-weight: bold; }
.Title { font-weight: bold; font-size: 18px; color: #cc3300; }
.Code { border: #8b4513 1px solid; padding-right: 5px; padding-left: 5px;color:
#000066; font-family: 'Courier New' , Monospace;background-color: #ff9933; }
Hi, my name is Hyun.Theoretically thinking, increasing the number of factor
shall enrich the translation model, thus generating better translation owing to
its discriminative capability. But in my case, it rather degrades the quality,
much worse than unfactored model. I doubt there are two reasons. In my case,
when I used the generation step, the system stops at runtime(decoding
session). But if I drop out the gen step, the decoder works..but with poor
quality. 2) Getting unexpected quality even after increasing the factor is
correlated with curse of dimensionality? I am studying pattern recognition
these days, and my brain says it may be due to the problem. Do I have to
have much more corpus when I
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
