def map_(cls, batches, host, port):
for sections in batches:
texts = [_.text for _ in sections]
data = post(cls, texts, host, port)
markups = list(parse(cls, texts, data))
for section, markup in strict_zip(sections, markups):
yield section.annotated(markup.spans)
def parse(texts, data):
for text, (chunks, tags) in strict_zip(texts, data):
# see patch_texts
if not text.strip():
spans = []
else:
tokens = list(find_tokens(chunks, text))
spans = list(bio_spans(tokens, tags))
yield DeeppavlovMarkup(text, spans)
def eval_score(tokens, guess, etalon, type):
spans = list(select_type_spans(guess, type))
guess_tags = list(spans_io(tokens, spans))
spans = list(select_type_spans(etalon, type))
etalon_tags = list(spans_io(tokens, spans))
score = Score()
for guess, etalon in strict_zip(guess_tags, etalon_tags):
guess, _ = parse_bio(guess)
etalon, _ = parse_bio(etalon)
if guess == I:
score.prec_total += 1
if etalon == O:
score.prec_errors += 1
if etalon == I:
score.recall_total += 1
if guess == O:
score.recall_errors += 1
return score
def eval_markups(guesses, etalons, types=TYPES):
scores = defaultdict(Score)
for guess, etalon in strict_zip(guesses, etalons):
for type, score in eval_markup(guess, etalon, types):
scores[type].update(score)
return dict(scores)