When I run mgiza, the a3.final and d3.final output files incorrectly have a
value of 100 for the length of every single target sentence. Does anyone
know how to fix this?

For example, just to test it, I ran a very small, toy sample where the
aligned corpora were as follows:
English:
I like big dogs
I like big cats
I like dogs
I like cats
cats like me
Indonesian:
saya suka anjing besar
saya suka kucing besar
saya suka anjing
saya suka kucing
kucing-kucing suka saya

Then, the resulting a3.final file looks like this:
1 1 3 100 1
2 2 3 100 1
1 3 3 100 0.333333
3 3 3 100 0.666667
1 1 4 100 1
2 2 4 100 1
4 3 4 100 1
3 4 4 100 1

As you can see, the 3rd column of this file is always 100, when it should
be 3 or 4.

Thanks so much!
Tom McCoy
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to