Hi! Yes, I managed to solve it.
During one my earlier attempts Giza++ had generated the vocabulary incorrectly for some reason. When the training perl script ran again it did not create the giza++ output files again as they existed earlier. This was the problem in my case - the faulty files that were being used. I realised that you should always clean up your work directory if you face an error as it can lead to some inconsistencies. Delete all files generated by Giza++/Moses and try again. I also removed some trailing non-significant terms from the vocab file (terms with very low frequency such as numeric values, symbols, incorrect spellings etc that creep in as part of the vocabulary). Dont think this would cause the problem as it would affect translations more than anything else but I think its a good idea to do so nevertheless. Ensure your training data is clean, sentences are aligned and lower cased(if applicable). Hope this helps. Regards, Danish On Tue, Dec 15, 2009 at 8:15 PM, Bill_Lang(Gmail) <[email protected]>wrote: > Hi friend, > I have encountered the same problem as yours, like in this link: > http://www.mail-archive.com/[email protected]/msg01646.html > > Before three months, I did the step-by-step guide for moses in > http://www.statmt.org/moses_steps.html > Then, everything is right. > > I do not know how to solve this problem. I want to know whether you have > solved it. If so, could you please tell me how to manage it? > > Best wishes; > Jun Lang > > >
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
