explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-03-12 07:15:48 +03:00
Code Issues Packages Projects Releases Wiki Activity

entities on doc_annotation, parse links and check their offsets against the entities. unit test works

Browse Source
This commit is contained in:
svlandeg 2020-06-12 15:47:20 +02:00
parent 3aed177a35
commit 880dccf93e
3 changed files with 115 additions and 53 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
spacy/errors.py
View File

@ -581,6 +581,10 @@ class Errors(object):
# TODO: fix numbering after merging develop into master
E983 = ("Each link annotation should refer to a dictionary with at most one "
"identifier mapping to 1.0, and all others to 0.0.")
E984 = ("The offsets of the annotations for 'links' need to refer exactly "
"to the offsets of the 'entities' annotations.")
E985 = ("The 'ent_iob' attribute of a Token should be an integer indexing "
"into {values}, but found {value}.")
E986 = ("Could not create any training batches: check your input. "

76
spacy/gold/new_example.pyx
View File

@ -85,12 +85,28 @@ cdef class NewExample:
return self.x.text
def _annot2array(strings, tok_annot, doc_annot):
def _annot2array(vocab, tok_annot, doc_annot):
attrs = []
values = []
for key, value in doc_annot.items():
if key == "entities":
words = tok_annot["ORTH"]
ent_iobs, ent_types = _parse_ner_tags(vocab, words, value)
tok_annot["ENT_IOB"] = ent_iobs
tok_annot["ENT_TYPE"] = ent_types
elif key == "links":
entities = doc_annot.get("entities", {})
if value and not entities:
raise ValueError(Errors.E984)
ent_kb_ids = _parse_links(vocab, words, value, entities)
tok_annot["ENT_KB_ID"] = ent_kb_ids
else:
raise ValueError(f"Unknown doc attribute: {key}")
for key, value in tok_annot.items():
if key not in IDS:
raise ValueError(f"Unknown attr: {key}")
raise ValueError(f"Unknown token attribute: {key}")
elif key == "ORTH":
pass
elif key == "HEAD":
@ -108,10 +124,8 @@ def _annot2array(strings, tok_annot, doc_annot):
raise ValueError(Errors.E985.format(values=iob_strings, value=values))
else:
attrs.append(key)
values.append([strings.add(v) for v in value])
# TODO: Calculate token.ent_kb_id from doc_annot["links"].
# We need to fix this and the doc.ents thing, both should be doc
# annotations.
values.append([vocab.strings.add(v) for v in value])
array = numpy.asarray(values, dtype="uint64")
return attrs, array.T
@ -129,8 +143,10 @@ def _fix_legacy_dict_data(predicted, example_dict):
for key, value in example_dict.items():
if key in ("token_annotation", "doc_annotation"):
pass
elif key in ("cats", "links"):
elif key in ("cats", "links") and value:
doc_dict[key] = value
elif key in ("ner", "entities") and value:
doc_dict["entities"] = value
else:
token_dict[key] = value
# Remap keys
@ -149,12 +165,6 @@ def _fix_legacy_dict_data(predicted, example_dict):
for key, value in old_token_dict.items():
if key in remapping:
token_dict[remapping[key]] = value
elif key in ("ner", "entities") and value:
# Arguably it would be smarter to put this in the doc annotation?
words = token_dict.get("words", [t.text for t in predicted])
ent_iobs, ent_types = _parse_ner_tags(predicted, words, value)
token_dict["ENT_IOB"] = ent_iobs
token_dict["ENT_TYPE"] = ent_types
else:
raise ValueError(f"Unknown attr: {key}")
return {
@ -163,16 +173,13 @@ def _fix_legacy_dict_data(predicted, example_dict):
}
def _parse_ner_tags(predicted, words, biluo_or_offsets):
def _parse_ner_tags(vocab, words, biluo_or_offsets):
if isinstance(biluo_or_offsets[0], (list, tuple)):
# Convert to biluo if necessary
# This is annoying but to convert the offsets we need a Doc
# that has the target tokenization.
reference = Doc(
predicted.vocab,
words=words
)
biluo = biluo_tags_from_offsets(predicted, biluo_or_offsets)
reference = Doc(vocab, words=words)
biluo = biluo_tags_from_offsets(reference, biluo_or_offsets)
else:
biluo = biluo_or_offsets
ent_iobs = []
@ -185,6 +192,37 @@ def _parse_ner_tags(predicted, words, biluo_or_offsets):
ent_types.append("")
return ent_iobs, ent_types
def _parse_links(vocab, words, links, entities):
reference = Doc(vocab, words=words)
starts = {token.idx: token.i for token in reference}
ends = {token.idx + len(token): token.i for token in reference}
ent_kb_ids = ["" for _ in reference]
entity_map = [(ent[0], ent[1]) for ent in entities]
# links annotations need to refer 1-1 to entity annotations - throw error otherwise
for index, annot_dict in links.items():
start_char, end_char = index
if (start_char, end_char) not in entity_map:
raise ValueError(Errors.E984)
for index, annot_dict in links.items():
true_kb_ids = []
for key, value in annot_dict.items():
if value == 1.0:
true_kb_ids.append(key)
if len(true_kb_ids) > 1:
raise ValueError(Errors.E983)
if len(true_kb_ids) == 1:
start_char, end_char = index
start_token = starts.get(start_char)
end_token = ends.get(end_char)
for i in range(start_token, end_token+1):
ent_kb_ids[i] = true_kb_ids[0]
return ent_kb_ids
class Example:
def get_aligned(self, field):

88
spacy/tests/test_new_example.py
View File

@ -41,33 +41,6 @@ def test_Example_from_dict_with_tags(annots):
assert token.tag_ == annots["tags"][i]
@pytest.mark.xfail(reason="TODO - fix")
@pytest.mark.parametrize(
"annots",
[
{
"words": ["I", "like", "New", "York", "and", "Berlin", "."],
"entities": [(7, 15, "LOC"), (20, 26, "LOC")],
}
],
)
def test_Example_from_dict_with_entities(annots):
vocab = Vocab()
predicted = Doc(vocab, words=annots["words"])
eg = Example.from_dict(predicted, annots)
assert len(list(eg.reference.ents)) == 2
assert eg.reference[0].ent_iob_ == "O"
assert eg.reference[1].ent_iob_ == "O"
assert eg.reference[2].ent_iob_ == "B"
assert eg.reference[3].ent_iob_ == "I"
assert eg.reference[4].ent_iob_ == "O"
assert eg.reference[5].ent_iob_ == "B"
assert eg.reference[6].ent_iob_ == "O"
assert eg.reference[2].ent_type_ == "LOC"
assert eg.reference[3].ent_type_ == "LOC"
assert eg.reference[5].ent_type_ == "LOC"
@pytest.mark.parametrize(
"annots",
[
@ -147,13 +120,39 @@ def test_Example_from_dict_with_cats(annots):
assert eg.reference.cats["cat3"] == 0.5
@pytest.mark.xfail(reason="TODO - fix")
@pytest.mark.parametrize(
"annots",
[
{
"words": ["Russ", "Cochran", "made", "reprints"],
"links": {(0, 12): {"Q7381115": 1.0, "Q2146908": 0.0}},
"words": ["I", "like", "New", "York", "and", "Berlin", "."],
"entities": [(7, 15, "LOC"), (20, 26, "LOC")],
}
],
)
def test_Example_from_dict_with_entities(annots):
vocab = Vocab()
predicted = Doc(vocab, words=annots["words"])
eg = Example.from_dict(predicted, annots)
assert len(list(eg.reference.ents)) == 2
assert eg.reference[0].ent_iob_ == "O"
assert eg.reference[1].ent_iob_ == "O"
assert eg.reference[2].ent_iob_ == "B"
assert eg.reference[3].ent_iob_ == "I"
assert eg.reference[4].ent_iob_ == "O"
assert eg.reference[5].ent_iob_ == "B"
assert eg.reference[6].ent_iob_ == "O"
assert eg.reference[2].ent_type_ == "LOC"
assert eg.reference[3].ent_type_ == "LOC"
assert eg.reference[5].ent_type_ == "LOC"
@pytest.mark.parametrize(
"annots",
[
{
"words": ["I", "like", "New", "York", "and", "Berlin", "."],
"entities": [(7, 15, "LOC"), (20, 26, "LOC")],
"links": {(7, 15): {"Q60": 1.0, "Q64": 0.0}, (20, 26): {"Q60": 0.0, "Q64": 1.0}},
}
],
)
@ -161,7 +160,28 @@ def test_Example_from_dict_with_links(annots):
vocab = Vocab()
predicted = Doc(vocab, words=annots["words"])
eg = Example.from_dict(predicted, annots)
assert eg.reference[0].ent_kb_id_ == "Q7381115"
assert eg.reference[1].ent_kb_id_ == "Q7381115"
assert eg.reference[2].ent_kb_id_ == ""
assert eg.reference[3].ent_kb_id_ == ""
assert eg.reference[0].ent_kb_id_ == ""
assert eg.reference[1].ent_kb_id_ == ""
assert eg.reference[2].ent_kb_id_ == "Q60"
assert eg.reference[3].ent_kb_id_ == "Q60"
assert eg.reference[4].ent_kb_id_ == ""
assert eg.reference[5].ent_kb_id_ == "Q64"
assert eg.reference[6].ent_kb_id_ == ""
@pytest.mark.parametrize(
"annots",
[
{
"words": ["I", "like", "New", "York", "and", "Berlin", "."],
"entities": [(7, 15, "LOC"), (20, 26, "LOC")],
"links": {(0, 1): {"Q7381115": 1.0, "Q2146908": 0.0}},
}
],
)
def test_Example_from_dict_with_links_invalid(annots):
vocab = Vocab()
predicted = Doc(vocab, words=annots["words"])
with pytest.raises(ValueError):
Example.from_dict(predicted, annots)