spaCy/spacy/tests/doc/test_retokenize_split.py

168 lines
6.0 KiB
Python
Raw Normal View History

# coding: utf-8
from __future__ import unicode_literals
import pytest
2019-02-14 18:56:38 +00:00
from spacy.vocab import Vocab
from spacy.tokens import Doc, Token
2019-02-14 18:56:38 +00:00
from ..util import get_doc
2019-02-24 13:14:11 +00:00
def test_doc_retokenize_split(en_vocab):
2019-02-15 11:56:51 +00:00
words = ["LosAngeles", "start", "."]
heads = [1, 1, 0]
2019-02-15 11:56:51 +00:00
doc = get_doc(en_vocab, words=words, heads=heads)
assert len(doc) == 3
assert len(str(doc)) == 19
2019-02-14 18:56:38 +00:00
assert doc[0].head.text == "start"
assert doc[1].head.text == "."
with doc.retokenize() as retokenizer:
retokenizer.split(
doc[0],
["Los", "Angeles"],
[(doc[0], 1), doc[1]],
attrs={
2019-02-17 11:22:07 +00:00
"tag": ["NNP"] * 2,
"lemma": ["Los", "Angeles"],
2019-02-17 11:22:07 +00:00
"ent_type": ["GPE"] * 2,
},
)
assert len(doc) == 4
2019-02-14 18:56:38 +00:00
assert doc[0].text == "Los"
assert doc[0].head.text == "Angeles"
assert doc[0].idx == 0
assert doc[1].idx == 3
2019-02-14 18:56:38 +00:00
assert doc[1].text == "Angeles"
assert doc[1].head.text == "start"
assert doc[2].text == "start"
assert doc[2].head.text == "."
assert doc[3].text == "."
assert doc[3].head.text == "."
assert len(str(doc)) == 19
2019-02-14 18:56:38 +00:00
2019-02-24 13:14:11 +00:00
def test_doc_retokenize_split_dependencies(en_vocab):
2019-02-15 11:56:51 +00:00
doc = Doc(en_vocab, words=["LosAngeles", "start", "."])
2019-02-14 18:56:38 +00:00
dep1 = doc.vocab.strings.add("amod")
dep2 = doc.vocab.strings.add("subject")
with doc.retokenize() as retokenizer:
2019-02-17 11:22:07 +00:00
retokenizer.split(
doc[0],
["Los", "Angeles"],
[(doc[0], 1), doc[1]],
attrs={"dep": [dep1, dep2]},
)
assert doc[0].dep == dep1
assert doc[1].dep == dep2
2019-02-24 13:14:11 +00:00
def test_doc_retokenize_split_heads_error(en_vocab):
2019-02-15 11:56:51 +00:00
doc = Doc(en_vocab, words=["LosAngeles", "start", "."])
2019-02-14 18:56:38 +00:00
# Not enough heads
with pytest.raises(ValueError):
with doc.retokenize() as retokenizer:
retokenizer.split(doc[0], ["Los", "Angeles"], [doc[1]])
2019-02-14 18:56:38 +00:00
# Too many heads
with pytest.raises(ValueError):
with doc.retokenize() as retokenizer:
retokenizer.split(doc[0], ["Los", "Angeles"], [doc[1], doc[1], doc[1]])
2019-02-24 13:14:11 +00:00
def test_doc_retokenize_spans_entity_split_iob():
# Test entity IOB stays consistent after merging
words = ["abc", "d", "e"]
doc = Doc(Vocab(), words=words)
2019-02-14 18:56:38 +00:00
doc.ents = [(doc.vocab.strings.add("ent-abcd"), 0, 2)]
assert doc[0].ent_iob_ == "B"
assert doc[1].ent_iob_ == "I"
with doc.retokenize() as retokenizer:
2019-02-17 11:22:07 +00:00
retokenizer.split(doc[0], ["a", "b", "c"], [(doc[0], 1), (doc[0], 2), doc[1]])
assert doc[0].ent_iob_ == "B"
assert doc[1].ent_iob_ == "I"
assert doc[2].ent_iob_ == "I"
assert doc[3].ent_iob_ == "I"
2019-02-14 18:56:38 +00:00
2019-02-24 13:14:11 +00:00
def test_doc_retokenize_spans_sentence_update_after_split(en_vocab):
2019-02-14 18:56:38 +00:00
# fmt: off
2019-02-15 11:56:51 +00:00
words = ["StewartLee", "is", "a", "stand", "up", "comedian", ".", "He",
"lives", "in", "England", "and", "loves", "JoePasquale", "."]
heads = [1, 0, 1, 2, -1, -4, -5, 1, 0, -1, -1, -3, -4, 1, -2]
2019-02-14 18:56:38 +00:00
deps = ["nsubj", "ROOT", "det", "amod", "prt", "attr", "punct", "nsubj",
"ROOT", "prep", "pobj", "cc", "conj", "compound", "punct"]
# fmt: on
2019-02-15 11:56:51 +00:00
doc = get_doc(en_vocab, words=words, heads=heads, deps=deps)
sent1, sent2 = list(doc.sents)
init_len = len(sent1)
init_len2 = len(sent2)
with doc.retokenize() as retokenizer:
2019-02-17 11:22:07 +00:00
retokenizer.split(
doc[0],
["Stewart", "Lee"],
[(doc[0], 1), doc[1]],
attrs={"dep": ["compound", "nsubj"]},
)
retokenizer.split(
doc[13],
["Joe", "Pasquale"],
[(doc[13], 1), doc[12]],
attrs={"dep": ["compound", "dobj"]},
)
sent1, sent2 = list(doc.sents)
assert len(sent1) == init_len + 1
assert len(sent2) == init_len2 + 1
2019-02-24 13:14:11 +00:00
def test_doc_retokenize_split_orths_mismatch(en_vocab):
"""Test that the regular retokenizer.split raises an error if the orths
don't match the original token text. There might still be a method that
allows this, but for the default use cases, merging and splitting should
always conform with spaCy's non-destructive tokenization policy. Otherwise,
it can lead to very confusing and unexpected results.
"""
doc = Doc(en_vocab, words=["LosAngeles", "start", "."])
with pytest.raises(ValueError):
with doc.retokenize() as retokenizer:
retokenizer.split(doc[0], ["L", "A"], [(doc[0], 0), (doc[0], 0)])
def test_doc_retokenize_split_extension_attrs(en_vocab):
Token.set_extension("a", default=False, force=True)
Token.set_extension("b", default="nothing", force=True)
doc = Doc(en_vocab, words=["LosAngeles", "start"])
with doc.retokenize() as retokenizer:
heads = [(doc[0], 1), doc[1]]
underscore = [{"a": True, "b": "1"}, {"b": "2"}]
attrs = {"lemma": ["los", "angeles"], "_": underscore}
retokenizer.split(doc[0], ["Los", "Angeles"], heads, attrs=attrs)
assert doc[0].lemma_ == "los"
2019-02-24 18:02:16 +00:00
assert doc[0]._.a is True
assert doc[0]._.b == "1"
assert doc[1].lemma_ == "angeles"
2019-02-24 18:02:16 +00:00
assert doc[1]._.a is False
assert doc[1]._.b == "2"
@pytest.mark.parametrize(
"underscore_attrs",
[
[{"a": "x"}, {}], # Overwriting getter without setter
[{"b": "x"}, {}], # Overwriting method
[{"c": "x"}, {}], # Overwriting nonexistent attribute
[{"a": "x"}, {"x": "x"}], # Combination
[{"a": "x", "x": "x"}, {"x": "x"}], # Combination
{"x": "x"}, # Not a list of dicts
],
)
def test_doc_retokenize_split_extension_attrs_invalid(en_vocab, underscore_attrs):
Token.set_extension("x", default=False, force=True)
Token.set_extension("a", getter=lambda x: x, force=True)
Token.set_extension("b", method=lambda x: x, force=True)
doc = Doc(en_vocab, words=["LosAngeles", "start"])
attrs = {"_": underscore_attrs}
with pytest.raises(ValueError):
with doc.retokenize() as retokenizer:
heads = [(doc[0], 1), doc[1]]
retokenizer.split(doc[0], ["Los", "Angeles"], heads, attrs=attrs)