def get_form_features(form, form_type, field_elems=None):
"""
Return a list of feature dicts, a dict per visible submittable
field in a <form> element.
"""
if field_elems is None:
field_elems = get_fields_to_annotate(form)
text_before, text_after = get_text_around_elems(form, field_elems)
res = [_elem_features(elem) for elem in field_elems]
for idx, elem_feat in enumerate(res):
if idx == 0:
elem_feat['is-first'] = True
if idx == len(res) - 1:
elem_feat['is-last'] = True
elem_feat['form-type'] = form_type
# get text before element
text = normalize(text_before[field_elems[idx]])
tokens = tokenize(text)[-6:]
elem_feat['text-before'] = token_ngrams(tokens, 1, 2)
# get text after element
text = normalize(text_after[field_elems[idx]])
tokens = tokenize(text)[:5]
elem_feat['text-after'] = token_ngrams(tokens, 1, 2)
elem_feat['bias'] = 1
return res
def get_form_features(form, form_type, field_elems=None):
"""
Return a list of feature dicts, a dict per visible submittable
field in a <form> element.
"""
if field_elems is None:
field_elems = get_fields_to_annotate(form)
text_before, text_after = get_text_around_elems(form, field_elems)
res = [_elem_features(elem) for elem in field_elems]
for idx, elem_feat in enumerate(res):
if idx == 0:
elem_feat['is-first'] = True
if idx == len(res)-1:
elem_feat['is-last'] = True
elem_feat['form-type'] = form_type
# get text before element
text = normalize(text_before[field_elems[idx]])
tokens = tokenize(text)[-6:]
elem_feat['text-before'] = token_ngrams(tokens, 1, 2)
# get text after element
text = normalize(text_after[field_elems[idx]])
tokens = tokenize(text)[:5]
elem_feat['text-after'] = token_ngrams(tokens, 1, 2)
elem_feat['bias'] = 1
return res
def keyword_text_relevancy(text: str,
pos_keywords: List[str],
neg_keywords: List[str],
max_ngram=1):
tokens = tokenize(text)
tokens = set(token_ngrams(tokens, 1, max_ngram))
def _score(keywords: List[str]) -> float:
s = sum(int(k in tokens) for k in keywords)
return _scale_relevancy(s, keywords)
pos_score = _score(pos_keywords)
neg_score = _score(neg_keywords)
return max(0, pos_score - 0.33 * neg_score)
def test_token_ngrams(seq, min_n, max_n, result):
assert token_ngrams(seq, min_n, max_n) == result
