hi avneesh
the source code for the counts should be in
score.cpp line 526
and
consolidate.coo line 275
the 1st count (56) is the counts of the target phrases, the 2nd (1) is
the count of the source phrases.
this email exchanges give some more info on it
http://thread.gmane.org/gmane.comp.nlp.moses.user/4639/focus=4646
On 31/10/2011 11:06, Avneesh Saluja wrote:
Hi, I just had a quick clarification question on how to parse the
phrase table. Say I have the following entry in my phrase table:
?? ? ||| if he ||| 0.0178571 0.00264583 1 0.244292 2.718 ||| ||| 56 1
On this line, we have the source phrase ||| target phrase ||| phrase
translation and lexical weighting probabilities ||| ||| counts
I traced through some of the code, and I saw that the first count
number is the "total count", and the second is the "original Count".
I'm not quite sure what each one refers to, but what I need are the
number of times the phrase pair (in this case [?? ?, if he]) occurs in
the training data.
Thanks,
Avneesh
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
