# Copyright 2024, Battelle Energy Alliance, LLC ALL RIGHTS RESERVED
from spacy.matcher import Matcher
from spacy.tokens import Span
from spacy.language import Language
# filter_spans is used to resolve the overlap issue in entities
# It gives primacy to longer spans (entities)
from spacy.util import filter_spans
import logging
[docs]
logger = logging.getLogger(__name__)
@Language.factory("simple_entity_matcher", default_config={"label": "ssc", "terms":[{"LOWER": "hello"}, {"IS_PUNCT": True}, {"LOWER": "world"}], "asSpan":True})
[docs]
def create_simple_matcher_component(nlp, name, label, terms, asSpan):
return SimpleEntityMatcher(nlp, label, terms, asSpan=asSpan)
[docs]
class SimpleEntityMatcher(object):
"""
How to use it:
.. code-block:: python
from SimpleEntityMatcher import SimpleEntityMatcher
nlp = spacy.load("en_core_web_sm")
terms = [{"LOWER": "hello"}, {"IS_PUNCT": True}, {"LOWER": "world"}]
pmatcher = SimpleEntityMatcher(nlp, 'ssc', terms)
doc = nlp("The shaft deflection is causing the safety cage to rattle. Pumps not experiencing enough flow for the pumps to keep the check valves open during test. Pump not experiencing enough flow during test. Shaft made noise. Vibration seems like it is coming from the shaft.")
updatedDoc = pmatcher(doc)
or:
.. code-block:: python
nlp.add_pipe('simple_entity_matcher', config={"label": "ssc", "terms":[{"LOWER": "hello"}, {"IS_PUNCT": True}, {"LOWER": "world"}], "asSpan":True})
newDoc = nlp(doc.text)
"""
def __init__(self, nlp, label, terms, asSpan=True, callback=None):
"""
Args:
nlp: spacy nlp model
label: str, the name/label for the patterns in terms
terms, list, the rules used to match the entities, for example,
terms = [{"LOWER": "hello"}, {"IS_PUNCT": True}, {"LOWER": "world"}]
"""
[docs]
self.name = 'simple_entity_matcher'
[docs]
self.matcher = Matcher(nlp.vocab)
if not isinstance(terms, list):
terms = [terms]
if not isinstance(terms[0], list):
terms = [terms]
self.matcher.add(label, terms, on_match=callback)
[docs]
def __call__(self, doc):
"""
Args:
doc: spacy.tokens.doc.Doc, the processed document using nlp pipelines
"""
matches = self.matcher(doc, as_spans=self.asSpan)
spans = []
if not self.asSpan:
for label, start, end in matches:
span = Span(doc, start, end, label=label)
spans.append(span)
else:
spans.extend(matches)
doc.ents = filter_spans(list(doc.ents)+spans)
return doc