Updated README: instead of decode, please run greedy (./greedy | ./grade) for our implementation.
There are two python programs here (-h for usage):
-decode translates input sentences from French to English.
-grade computes the model score of a translated sentence.
These commands work in a pipeline. For example:
> python decode | python grade
There is also a module:
-model.py implements very simple interfaces for language models
and translation models, so you don't have to.
You can finish the assignment without modifying this file at all. You should look at it if you need to understand the interface to the translation and language model.
The data directory contains files derived from the Canadian Hansards,
originally aligned by Ulrich Germann:
-input: French sentences to translate.
-tm: a phrase-based translation model. Each line is in the form:
French phrase ||| English phrase ||| log_10(translation_prob)
-lm: a trigram language model file in ARPA format.
log_10(ngram_prob) ngram log_10(backoff_prob)
The language model and translation model are computed from the data in the align directory, using alignments from the Berkeley aligner.