hi felipe

The decoder no longer read the word alignment information and the binary 
phrase table no longer carries that information.

there were problems with the implementation which made the decoder 
really slow and ate memory.

i have a new binary phrase table which does carry the information which 
you're welcome to use but you'll need to integrate it with the existing 
decoder.

Alternatively, roll back to before the alignment information was taken out

On 05/01/2010 19:49, Felipe Sánchez Martínez wrote:
> Hello all,
>
> I am trying to get the alignment information from a binary phrase table
> that was generated from a plain phrase table containing alignment
> information.
>
> I load the phrase table through the following piece of code:
> ------------------
> numScoreComponent=5;
> numInputScores=0;
> tableLimit=0;
> weightWP=0;
>
> input.push_back(0);
> output.push_back(0);
>
> weight.push_back(0); weight.push_back(0);
> weight.push_back(0); weight.push_back(0); weight.push_back(0);
>
> PhraseDictionaryTreeAdaptor *pd=new
> PhraseDictionaryTreeAdaptor(numScoreComponent, numInputScores);
>
> pd->DisableCache();
>
> if (!pd->Load(input, output, "/path/to/phrase/table", weight, tableLimit, 
> lmList, weightWP)) {
>      delete pd;
>      cerr<<"Error: Cannot load Moses binary phrase table, 
> path='"<<filePath<<"'\n";
>      exit(EXIT_FAILURE);
> }
> ------------------
>
> and I get the following error message:
>
> reading bin ttable
> size of OFF_T 8
> ERROR:You are asking binary phrase table without word alignments but the
> file do not exist. Please check if you had generated the correct phrase
> table without word alignment
> (phrase-table.binphr.srctree,phrase-table.binphr.tgtdata)
>
> ERROR:bin ttable was read in a wrong way
>
>
> When the binary phrase table does not contain any alignment information
> it works without problems.
>
> How should a binary phrase table with alignment information be loaded?
> Should I change some parameters? Which? Any help will be very valuable.
>
> Thank you very much in advance, and happy new year.
>
>    
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to