Hello everyone,
I'm trying to translate English to French. I am producing Lexicalized
Distortion models (with the typical bidirectional, FE, msd format) without
using the built-in scripts. I am feeding these to Moses and just want to
make sure my format and all is correct.
During training, A3 files are produced for with the following format:
# sentence pair (1) ...
french (target) sentence 1
english_word1 {french translation(s)} english_word2 {french translation(s)}
...
Likewise, there are "reverse" A3 files:
# sentence pair (1) ...
english (source) sentence 1
french_word1 {english translation(s)} french_word1 {english translations(s)}
*Question 1: *
Should my wbe-msd-FE-bidirectional distortion model have following format:
english ||| french ||| 6 scores
or
french ||| english ||| 6 scores
*Question 2:*
It seems that constructing such a model from the "regular" A3 files is
different than loading from the "reverse" A3 files. Which A3 files should
I be using?
Thanks!
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
