spaCy/examples/training/train_ner.py

113 lines
4.0 KiB
Python
Raw Normal View History

#!/usr/bin/env python
# coding: utf8
2017-11-01 02:43:22 +03:00
"""Example of training spaCy's named entity recognizer, starting off with an
existing model or a blank model.
For more details, see the documentation:
2017-11-07 14:00:43 +03:00
* Training: https://spacy.io/usage/training
* NER: https://spacy.io/usage/linguistic-features#named-entities
2017-11-07 03:22:30 +03:00
Compatible with: spaCy v2.0.0+
2019-03-16 16:15:49 +03:00
Last tested with: v2.1.0
"""
from __future__ import unicode_literals, print_function
2017-10-26 17:10:56 +03:00
import plac
import random
from pathlib import Path
import spacy
from spacy.util import minibatch, compounding
2017-01-27 14:27:10 +03:00
# training data
TRAIN_DATA = [
2018-12-17 15:44:38 +03:00
("Who is Shaka Khan?", {"entities": [(7, 17, "PERSON")]}),
("I like London and Berlin.", {"entities": [(7, 13, "LOC"), (18, 24, "LOC")]}),
]
2017-10-26 17:10:56 +03:00
@plac.annotations(
model=("Model name. Defaults to blank 'en' model.", "option", "m", str),
output_dir=("Optional output directory", "option", "o", Path),
2018-12-17 15:44:38 +03:00
n_iter=("Number of training iterations", "option", "n", int),
)
def main(model=None, output_dir=None, n_iter=100):
2017-10-26 17:10:56 +03:00
"""Load the model, set up the pipeline and train the entity recognizer."""
if model is not None:
nlp = spacy.load(model) # load existing spaCy model
print("Loaded model '%s'" % model)
else:
2018-12-17 15:44:38 +03:00
nlp = spacy.blank("en") # create blank Language class
print("Created blank 'en' model")
# create the built-in pipeline components and add them to the pipeline
2017-10-26 16:15:08 +03:00
# nlp.create_pipe works for built-ins that are registered with spaCy
2018-12-17 15:44:38 +03:00
if "ner" not in nlp.pipe_names:
ner = nlp.create_pipe("ner")
nlp.add_pipe(ner, last=True)
# otherwise, get it so we can add labels
else:
2018-12-17 15:44:38 +03:00
ner = nlp.get_pipe("ner")
# add labels
for _, annotations in TRAIN_DATA:
2018-12-17 15:44:38 +03:00
for ent in annotations.get("entities"):
ner.add_label(ent[2])
# get names of other pipes to disable them during training
pipe_exceptions = ["ner", "trf_wordpiecer", "trf_tok2vec"]
other_pipes = [pipe for pipe in nlp.pipe_names if pipe not in pipe_exceptions]
2017-10-27 01:31:30 +03:00
with nlp.disable_pipes(*other_pipes): # only train NER
# reset and initialize the weights randomly but only if we're
# training a new model
if model is None:
nlp.begin_training()
for itn in range(n_iter):
random.shuffle(TRAIN_DATA)
losses = {}
# batch up the examples using spaCy's minibatch
2018-12-17 15:44:38 +03:00
batches = minibatch(TRAIN_DATA, size=compounding(4.0, 32.0, 1.001))
for batch in batches:
texts, annotations = zip(*batch)
nlp.update(
texts, # batch of texts
annotations, # batch of annotations
drop=0.5, # dropout - make it harder to memorise data
2018-12-17 15:44:38 +03:00
losses=losses,
)
print("Losses", losses)
# test the trained model
for text, _ in TRAIN_DATA:
doc = nlp(text)
2018-12-17 15:44:38 +03:00
print("Entities", [(ent.text, ent.label_) for ent in doc.ents])
print("Tokens", [(t.text, t.ent_type_, t.ent_iob) for t in doc])
# save model to output directory
if output_dir is not None:
output_dir = Path(output_dir)
if not output_dir.exists():
output_dir.mkdir()
nlp.to_disk(output_dir)
print("Saved model to", output_dir)
# test the saved model
print("Loading from", output_dir)
2017-10-26 16:15:08 +03:00
nlp2 = spacy.load(output_dir)
for text, _ in TRAIN_DATA:
2017-10-26 16:15:08 +03:00
doc = nlp2(text)
2018-12-17 15:44:38 +03:00
print("Entities", [(ent.text, ent.label_) for ent in doc.ents])
print("Tokens", [(t.text, t.ent_type_, t.ent_iob) for t in doc])
2018-12-17 15:44:38 +03:00
if __name__ == "__main__":
2017-05-31 14:42:12 +03:00
plac.call(main)
# Expected output:
# Entities [('Shaka Khan', 'PERSON')]
# Tokens [('Who', '', 2), ('is', '', 2), ('Shaka', 'PERSON', 3),
# ('Khan', 'PERSON', 1), ('?', '', 2)]
# Entities [('London', 'LOC'), ('Berlin', 'LOC')]
# Tokens [('I', '', 2), ('like', '', 2), ('London', 'LOC', 3),
# ('and', '', 2), ('Berlin', 'LOC', 3), ('.', '', 2)]