These warnings aren't too serious.  The alignments should be fine.
Chris

On Thu, May 14, 2009 at 8:56 PM, Tom Hoar <[email protected]> wrote:
> During the train-factored-phrase-model.perl script of my
> character-by-character aligment, giza-- reported the errors below.
>
> Will this trained model be usable or do these errors indicate my model will
> unusable and I should restart?
>
> Thanks,
> Tom
>
>
> Model3: Iteration 1
> .
> 770000
> 780000
> PROBLEM: alignment is 0.
> NP 0.350144 APO 1 J:0.........
> WARNING: Hill Climbing yielded a zero score viterbi alignment for the
> following pair:
> AL(l:68,M:59)(a:.....................
> Source sentence length : 68 , target : 59
> 19 16 6 5 62..........
> 17 16 6 5 61..........
> PROBLEM: alignment is 0.
> NP 0.350144 APO 1 J:0.........
> WARNING: Hill Climbing yielded a zero score viterbi alignment for the
> following pair:
> AL(l:68,M:59)(a:.....................
> Source sentence length : 68 , target : 59
> 3 10 15 17 2 3........
> 3 10 14 18 2 23......
> 790000
> 800000
>
>
> Then, through Iterations 2, 3, & 4 it occationally reported these warnings:
>
> WARNING: already 41 iterations in hillclimb: xxxxxxx
> WARNING: already 42 iterations in hillclimb: xxxxxxx
> etc..
>
> Finally, in Model 4: Iteration 5, it reported:
>
> Reading more sentence pairs into memory ...
> 310000
> ^[OF320000
> 330000
> 340000
> 350000
> Reading more sentence pairs into memory ...
>
>
> _______________________________________________
> 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

Reply via email to