Hi,

the corpus is necessary to extract the actual paired words (or later phrases). 
In general, supply train-model with all parameters (even those that seem 
relevant only for specific steps) and restrict the processing to the steps 
needed.

Best luck,
  Ondrej.

"Kiran A Kate" <[email protected]> wrote:

>Hi,
>
>Thanks for the pointer. This was a problem since the script was just
>sleeping when there is a .lock file.
>Now, I can see an error reading a file, when I dug more into the script,
>seems that since I am starting with step 4, there is no corpus and
>somewhere the value of $__CORPUS is not initialized.
>Not sure if this is the problem but it tries to read a file .ur and
>obviously that does not exist.
>
>is the corpus really required?
>
>Thanks,
>Kiran.
>
>
>
>
>                                                                           
>             Ondrej Bojar                                                  
>             <[email protected]                                             
>             uni.cz>                                                    To 
>                                       Kiran A Kate/India/i...@ibmin,       
>             08/06/2010 17:18          [email protected]               
>                                                                        cc 
>                                                                           
>                                                                   Subject 
>                                       Re: [Moses-support] phrase table    
>                                       generation hangs                    
>                                                                           
>                                                                           
>                                                                           
>                                                                           
>                                                                           
>                                                                           
>
>
>
>
>Hi,
>
>I'm just guessing but it's likely that the script is expecting the file
>under a different name or at a different place. Due to the shell tools used
>e.g. to extract the file, it just keep waiting for input.
>
>Note that the two lexical translation dictionaries (1-1 word) are being
>prepared simultaneously so the progress messages are interleaved.
>
>Try adding debug prints at relevant places (search for the messages you're
>getting).
>
>Cheers,
>  Ondrej.
>
>"Kiran A Kate" <[email protected]> wrote:
>
>>
>>Hi,
>>
>>I am trying to use moses for phrase table generation based on the
>>alignments I have.
>>The alignments are manual and hence I intend to start the training from
>>step 4.
>>
>>
>$tooldir/moses/moses-scripts/scripts-20090926-1430/training/train-factored-phrase-model.perl
>
>> -scripts-root-dir $tooldir/moses/moses-scripts/scripts-20090926-1430
>>-root-dir work -f ur -e hi --first-step 4 -lm
>>0:3:/nls/p/798/ws/rb/out/hi.lm > work/training.out
>>
>>The alignment files hi-ur.A3.final.gz and ur-hi.A3.final.gz are there in
>>the default giza directories, i.e. work/giza.hi-ur and work/giza.ur-hi
>>respectively. (I created these files from the manual alignments in the
>>format as output by giza)
>>
>>This is the output I get:
>>$ Using
>>SCRIPTS_ROOTDIR: /nls/p/798/tools/moses/moses-scripts/scripts-20090926-1430
>
>>(4) generate lexical translation table 0-0 @ Tue Jun  8 06:56:42 EDT 2010
>>(4) [0-0] generate lexical translation table @ Tue Jun  8 06:56:42 EDT
>2010
>>Use of uninitialized value in concatenation (.) or string
>>at 
>>/nls/p/798/tools/moses/moses-scripts/scripts-20090926-1430/training/train-factored-phrase-model.perl
>
>> line 817.
>>(1.0.5) reducing factors to produce work/model/aligned.0.ur  @ Tue Jun  8
>>06:56:42 EDT 2010
>>(1.0.5) reducing factors to produce work/model/aligned.0.ur  @ Tue Jun  8
>>06:56:42 EDT 2010
>>
>>and it just hangs at that point. No error, no computation being performed.
>>
>>anybody has any idea, please help...
>>
>>Thanks,
>>Kiran.
>>
>>_______________________________________________
>>Moses-support mailing list
>>[email protected]
>>http://mailman.mit.edu/mailman/listinfo/moses-support
>
>
>--
>Ondrej Bojar
>http://www.cuni.cz/~obo
>


-- 
Ondrej Bojar
http://www.cuni.cz/~obo
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to