def train(self, parallel_corpus):
counts = Counts()
for aligned_sentence in parallel_corpus:
trg_sentence = aligned_sentence.words
src_sentence = [None] + aligned_sentence.mots
total_count = self.prob_all_alignments(src_sentence, trg_sentence)
for t in trg_sentence:
for s in src_sentence:
count = self.prob_alignment_point(s, t)
normalized_count = count / total_count[t]
counts.t_given_s[t][s] += normalized_count
counts.any_t_given_s[s] += normalized_count
self.maximize_lexical_translation_probabilities(counts)
def train(self, parallel_corpus):
counts = Counts()
for aligned_sentence in parallel_corpus:
trg_sentence = aligned_sentence.words
src_sentence = [None] + aligned_sentence.mots
# E step (a): Compute normalization factors to weigh counts
total_count = self.prob_all_alignments(src_sentence, trg_sentence)
# E step (b): Collect counts
for t in trg_sentence:
for s in src_sentence:
count = self.prob_alignment_point(s, t)
normalized_count = count / total_count[t]
counts.t_given_s[t][s] += normalized_count
counts.any_t_given_s[s] += normalized_count
# M step: Update probabilities with maximum likelihood estimate
self.maximize_lexical_translation_probabilities(counts)