mirror of
https://github.com/explosion/spaCy.git
synced 2024-09-21 03:19:13 +03:00
Improve GoldParse NER alignment (#5335)
Improve GoldParse NER alignment by including all cases where the start and end of the NER span can be aligned, regardless of internal tokenization differences. To do this, convert BILUO tags to character offsets, check start/end alignment with `doc.char_span()`, and assign the BILUO tags for the aligned spans. Alignment for `O/-` tags is handled through the one-to-one and multi alignments.
This commit is contained in:
adrianeboyd
GitHub
parent
481574cbc8
commit
84e06f9fb7
|
|
@ -110,6 +110,8 @@ class Warnings(object):
|
|||
W028 = ("Doc.from_array was called with a vector of type '{type}', "
|
||||
"but is expecting one of type 'uint64' instead. This may result "
|
||||
"in problems with the vocab further on in the pipeline.")
|
||||
W029 = ("Unable to align tokens with entities from character offsets. "
|
||||
"Discarding entity annotation for the text: {text}.")
|
||||
|
||||
|
||||
@add_codes
|
||||
|
|
|
|||
|
|
@ -648,6 +648,9 @@ cdef class GoldParse:
|
|||
# if self.lenght > 0, this is modified latter.
|
||||
self.orig_annot = []
|
||||
|
||||
# temporary doc for aligning entity annotation
|
||||
entdoc = None
|
||||
|
||||
# avoid allocating memory if the doc does not contain any tokens
|
||||
if self.length > 0:
|
||||
if words is None:
|
||||
|
|
@ -670,7 +673,25 @@ cdef class GoldParse:
|
|||
entities = [(ent if ent is not None else "-") for ent in entities]
|
||||
if not isinstance(entities[0], basestring):
|
||||
# Assume we have entities specified by character offset.
|
||||
entities = biluo_tags_from_offsets(doc, entities)
|
||||
# Create a temporary Doc corresponding to provided words
|
||||
# (to preserve gold tokenization) and text (to preserve
|
||||
# character offsets).
|
||||
entdoc_words, entdoc_spaces = util.get_words_and_spaces(words, doc.text)
|
||||
entdoc = Doc(doc.vocab, words=entdoc_words, spaces=entdoc_spaces)
|
||||
entdoc_entities = biluo_tags_from_offsets(entdoc, entities)
|
||||
# There may be some additional whitespace tokens in the
|
||||
# temporary doc, so check that the annotations align with
|
||||
# the provided words while building a list of BILUO labels.
|
||||
entities = []
|
||||
words_offset = 0
|
||||
for i in range(len(entdoc_words)):
|
||||
if words[i + words_offset] == entdoc_words[i]:
|
||||
entities.append(entdoc_entities[i])
|
||||
else:
|
||||
words_offset -= 1
|
||||
if len(entities) != len(words):
|
||||
user_warning(Warnings.W029.format(text=doc.text))
|
||||
entities = ["-" for _ in words]
|
||||
|
||||
# These are filled by the tagger/parser/entity recogniser
|
||||
self.c.tags = <int*>self.mem.alloc(len(doc), sizeof(int))
|
||||
|
|
@ -697,7 +718,8 @@ cdef class GoldParse:
|
|||
# If we under-segment, we'll have one predicted word that covers a
|
||||
# sequence of gold words.
|
||||
# If we "mis-segment", we'll have a sequence of predicted words covering
|
||||
# a sequence of gold words. That's many-to-many -- we don't do that.
|
||||
# a sequence of gold words. That's many-to-many -- we don't do that
|
||||
# except for NER spans where the start and end can be aligned.
|
||||
cost, i2j, j2i, i2j_multi, j2i_multi = align([t.orth_ for t in doc], words)
|
||||
|
||||
self.cand_to_gold = [(j if j >= 0 else None) for j in i2j]
|
||||
|
|
@ -720,7 +742,6 @@ cdef class GoldParse:
|
|||
self.tags[i] = tags[i2j_multi[i]]
|
||||
self.morphology[i] = morphology[i2j_multi[i]]
|
||||
is_last = i2j_multi[i] != i2j_multi.get(i+1)
|
||||
is_first = i2j_multi[i] != i2j_multi.get(i-1)
|
||||
# Set next word in multi-token span as head, until last
|
||||
if not is_last:
|
||||
self.heads[i] = i+1
|
||||
|
|
@ -730,30 +751,10 @@ cdef class GoldParse:
|
|||
if head_i:
|
||||
self.heads[i] = self.gold_to_cand[head_i]
|
||||
self.labels[i] = deps[i2j_multi[i]]
|
||||
# Now set NER...This is annoying because if we've split
|
||||
# got an entity word split into two, we need to adjust the
|
||||
# BILUO tags. We can't have BB or LL etc.
|
||||
# Case 1: O -- easy.
|
||||
ner_tag = entities[i2j_multi[i]]
|
||||
if ner_tag == "O":
|
||||
self.ner[i] = "O"
|
||||
# Case 2: U. This has to become a B I* L sequence.
|
||||
elif ner_tag.startswith("U-"):
|
||||
if is_first:
|
||||
self.ner[i] = ner_tag.replace("U-", "B-", 1)
|
||||
elif is_last:
|
||||
self.ner[i] = ner_tag.replace("U-", "L-", 1)
|
||||
else:
|
||||
self.ner[i] = ner_tag.replace("U-", "I-", 1)
|
||||
# Case 3: L. If not last, change to I.
|
||||
elif ner_tag.startswith("L-"):
|
||||
if is_last:
|
||||
self.ner[i] = ner_tag
|
||||
else:
|
||||
self.ner[i] = ner_tag.replace("L-", "I-", 1)
|
||||
# Case 4: I. Stays correct
|
||||
elif ner_tag.startswith("I-"):
|
||||
self.ner[i] = ner_tag
|
||||
# Assign O/- for many-to-one O/- NER tags
|
||||
if ner_tag in ("O", "-"):
|
||||
self.ner[i] = ner_tag
|
||||
else:
|
||||
self.words[i] = words[gold_i]
|
||||
self.tags[i] = tags[gold_i]
|
||||
|
|
@ -764,6 +765,39 @@ cdef class GoldParse:
|
|||
self.heads[i] = self.gold_to_cand[heads[gold_i]]
|
||||
self.labels[i] = deps[gold_i]
|
||||
self.ner[i] = entities[gold_i]
|
||||
# Assign O/- for one-to-many O/- NER tags
|
||||
for j, cand_j in enumerate(self.gold_to_cand):
|
||||
if cand_j is None:
|
||||
if j in j2i_multi:
|
||||
i = j2i_multi[j]
|
||||
ner_tag = entities[j]
|
||||
if ner_tag in ("O", "-"):
|
||||
self.ner[i] = ner_tag
|
||||
|
||||
# If there is entity annotation and some tokens remain unaligned,
|
||||
# align all entities at the character level to account for all
|
||||
# possible token misalignments within the entity spans
|
||||
if any([e not in ("O", "-") for e in entities]) and None in self.ner:
|
||||
# If the temporary entdoc wasn't created above, initialize it
|
||||
if not entdoc:
|
||||
entdoc_words, entdoc_spaces = util.get_words_and_spaces(words, doc.text)
|
||||
entdoc = Doc(doc.vocab, words=entdoc_words, spaces=entdoc_spaces)
|
||||
# Get offsets based on gold words and BILUO entities
|
||||
entdoc_offsets = offsets_from_biluo_tags(entdoc, entities)
|
||||
aligned_offsets = []
|
||||
aligned_spans = []
|
||||
# Filter offsets to identify those that align with doc tokens
|
||||
for offset in entdoc_offsets:
|
||||
span = doc.char_span(offset[0], offset[1])
|
||||
if span and not span.text.isspace():
|
||||
aligned_offsets.append(offset)
|
||||
aligned_spans.append(span)
|
||||
# Convert back to BILUO for doc tokens and assign NER for all
|
||||
# aligned spans
|
||||
biluo_tags = biluo_tags_from_offsets(doc, aligned_offsets, missing=None)
|
||||
for span in aligned_spans:
|
||||
for i in range(span.start, span.end):
|
||||
self.ner[i] = biluo_tags[i]
|
||||
|
||||
# Prevent whitespace that isn't within entities from being tagged as
|
||||
# an entity.
|
||||
|
|
|
|||
|
|
@ -6,6 +6,7 @@ from spacy.gold import spans_from_biluo_tags, GoldParse, iob_to_biluo
|
|||
from spacy.gold import GoldCorpus, docs_to_json, align
|
||||
from spacy.lang.en import English
|
||||
from spacy.tokens import Doc
|
||||
from spacy.util import get_words_and_spaces
|
||||
from .util import make_tempdir
|
||||
import pytest
|
||||
import srsly
|
||||
|
|
@ -59,6 +60,75 @@ def test_gold_biluo_misalign(en_vocab):
|
|||
assert tags == ["O", "O", "O", "-", "-", "-"]
|
||||
|
||||
|
||||
def test_gold_biluo_different_tokenization(en_vocab, en_tokenizer):
|
||||
# one-to-many
|
||||
words = ["I", "flew to", "San Francisco Valley", "."]
|
||||
spaces = [True, True, False, False]
|
||||
doc = Doc(en_vocab, words=words, spaces=spaces)
|
||||
entities = [(len("I flew to "), len("I flew to San Francisco Valley"), "LOC")]
|
||||
gp = GoldParse(
|
||||
doc,
|
||||
words=["I", "flew", "to", "San", "Francisco", "Valley", "."],
|
||||
entities=entities,
|
||||
)
|
||||
assert gp.ner == ["O", "O", "U-LOC", "O"]
|
||||
|
||||
# many-to-one
|
||||
words = ["I", "flew", "to", "San", "Francisco", "Valley", "."]
|
||||
spaces = [True, True, True, True, True, False, False]
|
||||
doc = Doc(en_vocab, words=words, spaces=spaces)
|
||||
entities = [(len("I flew to "), len("I flew to San Francisco Valley"), "LOC")]
|
||||
gp = GoldParse(
|
||||
doc, words=["I", "flew to", "San Francisco Valley", "."], entities=entities
|
||||
)
|
||||
assert gp.ner == ["O", "O", "O", "B-LOC", "I-LOC", "L-LOC", "O"]
|
||||
|
||||
# misaligned
|
||||
words = ["I flew", "to", "San Francisco", "Valley", "."]
|
||||
spaces = [True, True, True, False, False]
|
||||
doc = Doc(en_vocab, words=words, spaces=spaces)
|
||||
entities = [(len("I flew to "), len("I flew to San Francisco Valley"), "LOC")]
|
||||
gp = GoldParse(
|
||||
doc, words=["I", "flew to", "San", "Francisco Valley", "."], entities=entities,
|
||||
)
|
||||
assert gp.ner == ["O", "O", "B-LOC", "L-LOC", "O"]
|
||||
|
||||
# additional whitespace tokens in GoldParse words
|
||||
words, spaces = get_words_and_spaces(
|
||||
["I", "flew", "to", "San Francisco", "Valley", "."],
|
||||
"I flew to San Francisco Valley.",
|
||||
)
|
||||
doc = Doc(en_vocab, words=words, spaces=spaces)
|
||||
entities = [(len("I flew to "), len("I flew to San Francisco Valley"), "LOC")]
|
||||
gp = GoldParse(
|
||||
doc,
|
||||
words=["I", "flew", " ", "to", "San Francisco Valley", "."],
|
||||
entities=entities,
|
||||
)
|
||||
assert gp.ner == ["O", "O", "O", "O", "B-LOC", "L-LOC", "O"]
|
||||
|
||||
# from issue #4791
|
||||
data = (
|
||||
"I'll return the ₹54 amount",
|
||||
{
|
||||
"words": ["I", "'ll", "return", "the", "₹", "54", "amount",],
|
||||
"entities": [(16, 19, "MONEY")],
|
||||
},
|
||||
)
|
||||
gp = GoldParse(en_tokenizer(data[0]), **data[1])
|
||||
assert gp.ner == ["O", "O", "O", "O", "U-MONEY", "O"]
|
||||
|
||||
data = (
|
||||
"I'll return the $54 amount",
|
||||
{
|
||||
"words": ["I", "'ll", "return", "the", "$", "54", "amount",],
|
||||
"entities": [(16, 19, "MONEY")],
|
||||
},
|
||||
)
|
||||
gp = GoldParse(en_tokenizer(data[0]), **data[1])
|
||||
assert gp.ner == ["O", "O", "O", "O", "B-MONEY", "L-MONEY", "O"]
|
||||
|
||||
|
||||
def test_roundtrip_offsets_biluo_conversion(en_tokenizer):
|
||||
text = "I flew to Silicon Valley via London."
|
||||
biluo_tags = ["O", "O", "O", "B-LOC", "L-LOC", "O", "U-GPE", "O"]
|
||||
|
|
|
|||
|
|
@ -758,7 +758,7 @@ def get_serialization_exclude(serializers, exclude, kwargs):
|
|||
|
||||
|
||||
def get_words_and_spaces(words, text):
|
||||
if "".join("".join(words).split())!= "".join(text.split()):
|
||||
if "".join("".join(words).split()) != "".join(text.split()):
|
||||
raise ValueError(Errors.E194.format(text=text, words=words))
|
||||
text_words = []
|
||||
text_spaces = []
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user