Using lexicon probabilities

[Chloe Braud]

Hi Graham and pialign users,

I'm not sure to understand the -le2f and -lf2e options, do they correspond to bi-lingual lexicons where words and their translations are associated with a probability? If so, what is the expected format for these files?

Thanks,

[Graham Neubig]

Hi Chloe,

These are in space-separated format where it's

word_f word_e probability

Graham

--
You received this message because you are subscribed to the Google Groups "pialign-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pialign-users+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Chloe Braud]

Hi Graham,

Thanks a lot for your answer! That actually raises new questions:

1- Is it possible to have composed words (that may contain a space) in the lexicon? 

2- I'm not sure to understand how this lexicon is used by the model: is it used to initialize the conditional probabilities at the word level?

3- Is it possible to give a (seed) phrase table as input?

All the best,

Le mardi 14 février 2017 18:47:42 UTC+1, Graham Neubig a écrit :

Hi Chloe,

These are in space-separated format where it's

word_f word_e probability

Graham

On Tue, Feb 14, 2017 at 9:59 AM, Chloe Braud <chloe...@gmail.com> wrote:

Hi Graham and pialign users,

I'm not sure to understand the -le2f and -lf2e options, do they correspond to bi-lingual lexicons where words and their translations are associated with a probability? If so, what is the expected format for these files?

Thanks,

--
You received this message because you are subscribed to the Google Groups "pialign-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pialign-user...@googlegroups.com.