def _extend_associations_with(associations: List[Association],
words: List[str],
model: KeyedVectors) -> List[Association]:
valid_pos_tags_for_associations = associations_config.VALID_POS_TAGS_FOR_ASSOCIATIONS
valid_pos_tags_for_associated_words = associations_config.VALID_POS_TAGS_FOR_ASSOCIATED_WORDS
result = []
for association in associations:
association_word_with_tag = _add_pos_tag(
association.association_word,
_get_most_likely_pos_tag(association.association_word,
valid_pos_tags_for_associations, model))
for word in words:
if association.has_associated_word(word):
continue
word_with_tag = _add_pos_tag(
word,
_get_most_likely_pos_tag(word,
valid_pos_tags_for_associated_words,
model))
similarity = model.similarity(association_word_with_tag,
word_with_tag)
if similarity < associations_config.MIN_SIMILARITY_SCORE:
continue
new_association = association.copy()
new_association.add_associated_word((word, similarity))
result.append(new_association)
return result
def _add_rival_words_and_filter_associations(
associations: List[Association],
rival_words_with_coefficients: List[Tuple[str, float]],
model: KeyedVectors) -> List[Association]:
result: List[Association] = []
valid_pos_tags_for_associations = associations_config.VALID_POS_TAGS_FOR_ASSOCIATIONS
valid_pos_tags_for_associated_words = associations_config.VALID_POS_TAGS_FOR_ASSOCIATED_WORDS
for association in associations:
min_associated_word_score = min(association.associated_word_scores)
association_word_with_tag = _add_pos_tag(
association.association_word,
_get_most_likely_pos_tag(association.association_word,
valid_pos_tags_for_associations, model))
association_is_too_dangerous = False
for rival_word, rival_word_coefficient in rival_words_with_coefficients:
rival_word_with_pos_tag = _add_pos_tag(
rival_word,
_get_most_likely_pos_tag(rival_word,
valid_pos_tags_for_associated_words,
model))
rival_word_score = model.similarity(association_word_with_tag,
rival_word_with_pos_tag)
scaled_rival_word_score = rival_word_coefficient * rival_word_score
if scaled_rival_word_score > min_associated_word_score:
association_is_too_dangerous = True
break
association.add_rival_word((rival_word, float(rival_word_score)))
if not association_is_too_dangerous:
result.append(association)
return result
