def score_model(vocab, tagger, parser, gold_docs, verbose=False):
scorer = Scorer()
for _, gold_doc in gold_docs:
for (ids, words, tags, heads, deps, entities), _ in gold_doc:
doc = Doc(vocab, words=words)
tagger(doc)
parser(doc)
PseudoProjectivity.deprojectivize(doc)
gold = GoldParse(doc, tags=tags, heads=heads, deps=deps)
scorer.score(doc, gold, verbose=verbose)
return scorer
def score_model(vocab, tagger, parser, gold_docs, verbose=False):
scorer = Scorer()
for _, gold_doc in gold_docs:
for (ids, words, tags, heads, deps, entities), _ in gold_doc:
doc = Doc(vocab, words=words)
tagger(doc)
parser(doc)
PseudoProjectivity.deprojectivize(doc)
gold = GoldParse(doc, tags=tags, heads=heads, deps=deps)
scorer.score(doc, gold, verbose=verbose)
return scorer
def deprojectivize(proj_heads, deco_labels, EN):
slen = len(proj_heads)
sent = EN.tokenizer.tokens_from_list(["whatever"] * slen)
rel_proj_heads = [head - i for i, head in enumerate(proj_heads)]
labelids = [EN.vocab.strings[label] for label in deco_labels]
pairs = list(zip(rel_proj_heads, labelids))
parse = numpy.asarray(pairs, dtype=numpy.int32)
sent.from_array([HEAD, DEP], parse)
PseudoProjectivity.deprojectivize(sent)
parse = sent.to_array([HEAD, DEP])
deproj_heads = [i + head for i, head in enumerate(parse[:, 0])]
undeco_labels = [EN.vocab.strings[int(labelid)] for labelid in parse[:, 1]]
return deproj_heads, undeco_labels
def deprojectivize(proj_heads, deco_labels, EN):
slen = len(proj_heads)
sent = EN.tokenizer.tokens_from_list(['whatever'] * slen)
rel_proj_heads = [head - i for i, head in enumerate(proj_heads)]
labelids = [EN.vocab.strings[label] for label in deco_labels]
pairs = list(zip(rel_proj_heads, labelids))
parse = numpy.asarray(pairs, dtype=numpy.int32)
sent.from_array([HEAD, DEP], parse)
PseudoProjectivity.deprojectivize(sent)
parse = sent.to_array([HEAD, DEP])
deproj_heads = [i + head for i, head in enumerate(parse[:, 0])]
undeco_labels = [EN.vocab.strings[int(labelid)] for labelid in parse[:, 1]]
return deproj_heads, undeco_labels
