mirror of
https://github.com/explosion/spaCy.git
synced 2024-11-10 19:57:17 +03:00
830dcca367
When the default `max_length` is not set and there are longer training documents, it can be difficult to train and evaluate the span finder due to memory limits and the time it takes to evaluate a huge number of predicted spans.
336 lines
12 KiB
Python
336 lines
12 KiB
Python
from typing import Any, Callable, Dict, Iterable, List, Optional, Tuple
|
|
|
|
from thinc.api import Config, Model, Optimizer, set_dropout_rate
|
|
from thinc.types import Floats2d
|
|
|
|
from ..errors import Errors
|
|
from ..language import Language
|
|
from ..scorer import Scorer
|
|
from ..tokens import Doc, Span
|
|
from ..training import Example
|
|
from ..util import registry
|
|
from .spancat import DEFAULT_SPANS_KEY
|
|
from .trainable_pipe import TrainablePipe
|
|
|
|
span_finder_default_config = """
|
|
[model]
|
|
@architectures = "spacy.SpanFinder.v1"
|
|
|
|
[model.scorer]
|
|
@layers = "spacy.LinearLogistic.v1"
|
|
nO = 2
|
|
|
|
[model.tok2vec]
|
|
@architectures = "spacy.Tok2Vec.v2"
|
|
|
|
[model.tok2vec.embed]
|
|
@architectures = "spacy.MultiHashEmbed.v2"
|
|
width = 96
|
|
rows = [5000, 1000, 2500, 1000]
|
|
attrs = ["NORM", "PREFIX", "SUFFIX", "SHAPE"]
|
|
include_static_vectors = false
|
|
|
|
[model.tok2vec.encode]
|
|
@architectures = "spacy.MaxoutWindowEncoder.v2"
|
|
width = ${model.tok2vec.embed.width}
|
|
window_size = 1
|
|
maxout_pieces = 3
|
|
depth = 4
|
|
"""
|
|
|
|
DEFAULT_SPAN_FINDER_MODEL = Config().from_str(span_finder_default_config)["model"]
|
|
|
|
|
|
@Language.factory(
|
|
"span_finder",
|
|
assigns=["doc.spans"],
|
|
default_config={
|
|
"threshold": 0.5,
|
|
"model": DEFAULT_SPAN_FINDER_MODEL,
|
|
"spans_key": DEFAULT_SPANS_KEY,
|
|
"max_length": 25,
|
|
"min_length": None,
|
|
"scorer": {"@scorers": "spacy.span_finder_scorer.v1"},
|
|
},
|
|
default_score_weights={
|
|
f"spans_{DEFAULT_SPANS_KEY}_f": 1.0,
|
|
f"spans_{DEFAULT_SPANS_KEY}_p": 0.0,
|
|
f"spans_{DEFAULT_SPANS_KEY}_r": 0.0,
|
|
},
|
|
)
|
|
def make_span_finder(
|
|
nlp: Language,
|
|
name: str,
|
|
model: Model[Iterable[Doc], Floats2d],
|
|
spans_key: str,
|
|
threshold: float,
|
|
max_length: Optional[int],
|
|
min_length: Optional[int],
|
|
scorer: Optional[Callable],
|
|
) -> "SpanFinder":
|
|
"""Create a SpanFinder component. The component predicts whether a token is
|
|
the start or the end of a potential span.
|
|
|
|
model (Model[List[Doc], Floats2d]): A model instance that
|
|
is given a list of documents and predicts a probability for each token.
|
|
spans_key (str): Key of the doc.spans dict to save the spans under. During
|
|
initialization and training, the component will look for spans on the
|
|
reference document under the same key.
|
|
threshold (float): Minimum probability to consider a prediction positive.
|
|
max_length (Optional[int]): Maximum length of the produced spans, defaults
|
|
to None meaning unlimited length.
|
|
min_length (Optional[int]): Minimum length of the produced spans, defaults
|
|
to None meaning shortest span length is 1.
|
|
scorer (Optional[Callable]): The scoring method. Defaults to
|
|
Scorer.score_spans for the Doc.spans[spans_key] with overlapping
|
|
spans allowed.
|
|
"""
|
|
return SpanFinder(
|
|
nlp,
|
|
model=model,
|
|
threshold=threshold,
|
|
name=name,
|
|
scorer=scorer,
|
|
max_length=max_length,
|
|
min_length=min_length,
|
|
spans_key=spans_key,
|
|
)
|
|
|
|
|
|
@registry.scorers("spacy.span_finder_scorer.v1")
|
|
def make_span_finder_scorer():
|
|
return span_finder_score
|
|
|
|
|
|
def span_finder_score(examples: Iterable[Example], **kwargs) -> Dict[str, Any]:
|
|
kwargs = dict(kwargs)
|
|
attr_prefix = "spans_"
|
|
key = kwargs["spans_key"]
|
|
kwargs.setdefault("attr", f"{attr_prefix}{key}")
|
|
kwargs.setdefault(
|
|
"getter", lambda doc, key: doc.spans.get(key[len(attr_prefix) :], [])
|
|
)
|
|
kwargs.setdefault("has_annotation", lambda doc: key in doc.spans)
|
|
kwargs.setdefault("allow_overlap", True)
|
|
kwargs.setdefault("labeled", False)
|
|
scores = Scorer.score_spans(examples, **kwargs)
|
|
scores.pop(f"{kwargs['attr']}_per_type", None)
|
|
return scores
|
|
|
|
|
|
def _char_indices(span: Span) -> Tuple[int, int]:
|
|
start = span[0].idx
|
|
end = span[-1].idx + len(span[-1])
|
|
return start, end
|
|
|
|
|
|
class SpanFinder(TrainablePipe):
|
|
"""Pipeline that learns span boundaries.
|
|
|
|
DOCS: https://spacy.io/api/spanfinder
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
nlp: Language,
|
|
model: Model[Iterable[Doc], Floats2d],
|
|
name: str = "span_finder",
|
|
*,
|
|
spans_key: str = DEFAULT_SPANS_KEY,
|
|
threshold: float = 0.5,
|
|
max_length: Optional[int] = None,
|
|
min_length: Optional[int] = None,
|
|
scorer: Optional[Callable] = span_finder_score,
|
|
) -> None:
|
|
"""Initialize the span finder.
|
|
model (thinc.api.Model): The Thinc Model powering the pipeline
|
|
component.
|
|
name (str): The component instance name, used to add entries to the
|
|
losses during training.
|
|
threshold (float): Minimum probability to consider a prediction
|
|
positive.
|
|
scorer (Optional[Callable]): The scoring method.
|
|
spans_key (str): Key of the doc.spans dict to save the spans under.
|
|
During initialization and training, the component will look for
|
|
spans on the reference document under the same key.
|
|
max_length (Optional[int]): Maximum length of the produced spans,
|
|
defaults to None meaning unlimited length.
|
|
min_length (Optional[int]): Minimum length of the produced spans,
|
|
defaults to None meaning shortest span length is 1.
|
|
|
|
DOCS: https://spacy.io/api/spanfinder#init
|
|
"""
|
|
self.vocab = nlp.vocab
|
|
if (max_length is not None and max_length < 1) or (
|
|
min_length is not None and min_length < 1
|
|
):
|
|
raise ValueError(
|
|
Errors.E1053.format(min_length=min_length, max_length=max_length)
|
|
)
|
|
self.model = model
|
|
self.name = name
|
|
self.scorer = scorer
|
|
self.cfg: Dict[str, Any] = {
|
|
"min_length": min_length,
|
|
"max_length": max_length,
|
|
"threshold": threshold,
|
|
"spans_key": spans_key,
|
|
}
|
|
|
|
def predict(self, docs: Iterable[Doc]):
|
|
"""Apply the pipeline's model to a batch of docs, without modifying
|
|
them.
|
|
|
|
docs (Iterable[Doc]): The documents to predict.
|
|
RETURNS: The models prediction for each document.
|
|
|
|
DOCS: https://spacy.io/api/spanfinder#predict
|
|
"""
|
|
scores = self.model.predict(docs)
|
|
return scores
|
|
|
|
def set_annotations(self, docs: Iterable[Doc], scores: Floats2d) -> None:
|
|
"""Modify a batch of Doc objects, using pre-computed scores.
|
|
docs (Iterable[Doc]): The documents to modify.
|
|
scores: The scores to set, produced by SpanFinder predict method.
|
|
|
|
DOCS: https://spacy.io/api/spanfinder#set_annotations
|
|
"""
|
|
offset = 0
|
|
for i, doc in enumerate(docs):
|
|
doc.spans[self.cfg["spans_key"]] = []
|
|
starts = []
|
|
ends = []
|
|
doc_scores = scores[offset : offset + len(doc)]
|
|
|
|
for token, token_score in zip(doc, doc_scores):
|
|
if token_score[0] >= self.cfg["threshold"]:
|
|
starts.append(token.i)
|
|
if token_score[1] >= self.cfg["threshold"]:
|
|
ends.append(token.i)
|
|
|
|
for start in starts:
|
|
for end in ends:
|
|
span_length = end + 1 - start
|
|
if span_length < 1:
|
|
continue
|
|
if (
|
|
self.cfg["min_length"] is None
|
|
or self.cfg["min_length"] <= span_length
|
|
) and (
|
|
self.cfg["max_length"] is None
|
|
or span_length <= self.cfg["max_length"]
|
|
):
|
|
doc.spans[self.cfg["spans_key"]].append(doc[start : end + 1])
|
|
offset += len(doc)
|
|
|
|
def update(
|
|
self,
|
|
examples: Iterable[Example],
|
|
*,
|
|
drop: float = 0.0,
|
|
sgd: Optional[Optimizer] = None,
|
|
losses: Optional[Dict[str, float]] = None,
|
|
) -> Dict[str, float]:
|
|
"""Learn from a batch of documents and gold-standard information,
|
|
updating the pipe's model. Delegates to predict and get_loss.
|
|
examples (Iterable[Example]): A batch of Example objects.
|
|
drop (float): The dropout rate.
|
|
sgd (Optional[thinc.api.Optimizer]): The optimizer.
|
|
losses (Optional[Dict[str, float]]): Optional record of the loss during
|
|
training. Updated using the component name as the key.
|
|
RETURNS (Dict[str, float]): The updated losses dictionary.
|
|
|
|
DOCS: https://spacy.io/api/spanfinder#update
|
|
"""
|
|
if losses is None:
|
|
losses = {}
|
|
losses.setdefault(self.name, 0.0)
|
|
predicted = [eg.predicted for eg in examples]
|
|
set_dropout_rate(self.model, drop)
|
|
scores, backprop_scores = self.model.begin_update(predicted)
|
|
loss, d_scores = self.get_loss(examples, scores)
|
|
backprop_scores(d_scores)
|
|
if sgd is not None:
|
|
self.finish_update(sgd)
|
|
losses[self.name] += loss
|
|
return losses
|
|
|
|
def get_loss(self, examples, scores) -> Tuple[float, Floats2d]:
|
|
"""Find the loss and gradient of loss for the batch of documents and
|
|
their predicted scores.
|
|
examples (Iterable[Examples]): The batch of examples.
|
|
scores: Scores representing the model's predictions.
|
|
RETURNS (Tuple[float, Floats2d]): The loss and the gradient.
|
|
|
|
DOCS: https://spacy.io/api/spanfinder#get_loss
|
|
"""
|
|
truths, masks = self._get_aligned_truth_scores(examples, self.model.ops)
|
|
d_scores = scores - self.model.ops.asarray2f(truths)
|
|
d_scores *= masks
|
|
loss = float((d_scores**2).sum())
|
|
return loss, d_scores
|
|
|
|
def _get_aligned_truth_scores(self, examples, ops) -> Tuple[Floats2d, Floats2d]:
|
|
"""Align scores of the predictions to the references for calculating
|
|
the loss.
|
|
"""
|
|
truths = []
|
|
masks = []
|
|
for eg in examples:
|
|
if eg.x.text != eg.y.text:
|
|
raise ValueError(Errors.E1054.format(component="span_finder"))
|
|
n_tokens = len(eg.predicted)
|
|
truth = ops.xp.zeros((n_tokens, 2), dtype="float32")
|
|
mask = ops.xp.ones((n_tokens, 2), dtype="float32")
|
|
if self.cfg["spans_key"] in eg.reference.spans:
|
|
for span in eg.reference.spans[self.cfg["spans_key"]]:
|
|
ref_start_char, ref_end_char = _char_indices(span)
|
|
pred_span = eg.predicted.char_span(
|
|
ref_start_char, ref_end_char, alignment_mode="expand"
|
|
)
|
|
pred_start_char, pred_end_char = _char_indices(pred_span)
|
|
start_match = pred_start_char == ref_start_char
|
|
end_match = pred_end_char == ref_end_char
|
|
if start_match:
|
|
truth[pred_span[0].i, 0] = 1
|
|
else:
|
|
mask[pred_span[0].i, 0] = 0
|
|
if end_match:
|
|
truth[pred_span[-1].i, 1] = 1
|
|
else:
|
|
mask[pred_span[-1].i, 1] = 0
|
|
truths.append(truth)
|
|
masks.append(mask)
|
|
truths = ops.xp.concatenate(truths, axis=0)
|
|
masks = ops.xp.concatenate(masks, axis=0)
|
|
return truths, masks
|
|
|
|
def initialize(
|
|
self,
|
|
get_examples: Callable[[], Iterable[Example]],
|
|
*,
|
|
nlp: Optional[Language] = None,
|
|
) -> None:
|
|
"""Initialize the pipe for training, using a representative set
|
|
of data examples.
|
|
get_examples (Callable[[], Iterable[Example]]): Function that
|
|
returns a representative sample of gold-standard Example objects.
|
|
nlp (Optional[Language]): The current nlp object the component is part
|
|
of.
|
|
|
|
DOCS: https://spacy.io/api/spanfinder#initialize
|
|
"""
|
|
subbatch: List[Example] = []
|
|
|
|
for eg in get_examples():
|
|
if len(subbatch) < 10:
|
|
subbatch.append(eg)
|
|
|
|
if subbatch:
|
|
docs = [eg.reference for eg in subbatch]
|
|
Y, _ = self._get_aligned_truth_scores(subbatch, self.model.ops)
|
|
self.model.initialize(X=docs, Y=Y)
|
|
else:
|
|
self.model.initialize()
|