Description
BERT Model
with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks.
bert_large_token_classifier_ontonote is a fine-tuned BERT model that is ready to use for Named Entity Recognition and achieves state-of-the-art performance for the NER task. This model has been trained to recognize four types of entities: CARDINAL, DATE, EVENT, FAC, GPE, LANGUAGE, LAW, LOC, MONEY, NORP, ORDINAL, ORG, PERCENT, PERSON, PRODUCT, QUANTITY, TIME, and WORK_OF_ART.
We used TFBertForTokenClassification to train this model and used BertForTokenClassification
annotator in Spark NLP 🚀 for prediction at scale!
Predicted Entities
CARDINAL
, DATE
, EVENT
, FAC
, GPE
, LANGUAGE
, LAW
, LOC
, MONEY
, NORP
, ORDINAL
, ORG
, PERCENT
, PERSON
, PRODUCT
, QUANTITY
, TIME
, WORK_OF_ART
How to use
document_assembler = DocumentAssembler() \
.setInputCol('text') \
.setOutputCol('document')
tokenizer = Tokenizer() \
.setInputCols(['document']) \
.setOutputCol('token')
tokenClassifier = BertForTokenClassification \
.pretrained('bert_large_token_classifier_ontonote', 'en') \
.setInputCols(['token', 'document']) \
.setOutputCol('ner') \
.setCaseSensitive(True) \
.setMaxSentenceLength(512)
# since output column is IOB/IOB2 style, NerConverter can extract entities
ner_converter = NerConverter() \
.setInputCols(['document', 'token', 'ner']) \
.setOutputCol('entities')
pipeline = Pipeline(stages=[
document_assembler,
tokenizer,
tokenClassifier,
ner_converter
])
example = spark.createDataFrame([['My name is John!']]).toDF("text")
result = pipeline.fit(example).transform(example)
val document_assembler = DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val tokenizer = Tokenizer()
.setInputCols("document")
.setOutputCol("token")
val tokenClassifier = BertForTokenClassification.pretrained("bert_large_token_classifier_ontonote", "en")
.setInputCols("document", "token")
.setOutputCol("ner")
.setCaseSensitive(true)
.setMaxSentenceLength(512)
// since output column is IOB/IOB2 style, NerConverter can extract entities
val ner_converter = NerConverter()
.setInputCols("document", "token", "ner")
.setOutputCol("entities")
val pipeline = new Pipeline().setStages(Array(document_assembler, tokenizer, tokenClassifier, ner_converter))
val example = Seq.empty["My name is John!"].toDS.toDF("text")
val result = pipeline.fit(example).transform(example)
import nlu
nlu.load("en.classify.token_bert.large_ontonote").predict("""My name is John!""")
Results
+------------------------------------------------------------------------------------+
|result |
+------------------------------------------------------------------------------------+
|[B-PERSON, I-PERSON, O, O, O, B-LOC, O, O, O, B-LOC, O, O, O, O, B-PERSON, O, O, O, O, B-LOC]|
+------------------------------------------------------------------------------------+
Model Information
Model Name: | bert_large_token_classifier_ontonote |
Compatibility: | Spark NLP 3.2.0+ |
License: | Open Source |
Edition: | Official |
Input Labels: | [token, document] |
Output Labels: | [ner] |
Language: | en |
Case sensitive: | true |
Max sentense length: | 512 |
Data Source
https://catalog.ldc.upenn.edu/LDC2013T19
Benchmarking
Test:
precision recall f1-score support
B-CARDINAL 0.86 0.86 0.86 935
B-DATE 0.88 0.89 0.88 1602
B-EVENT 0.76 0.67 0.71 63
B-FAC 0.77 0.84 0.81 135
B-GPE 0.98 0.92 0.95 2240
B-LANGUAGE 0.79 0.68 0.73 22
B-LAW 0.77 0.68 0.72 40
B-LOC 0.73 0.82 0.78 179
B-MONEY 0.90 0.89 0.89 314
B-NORP 0.94 0.96 0.95 841
B-ORDINAL 0.82 0.91 0.87 195
B-ORG 0.90 0.91 0.91 1795
B-PERCENT 0.94 0.93 0.94 349
B-PERSON 0.95 0.96 0.95 1988
B-PRODUCT 0.79 0.80 0.80 76
B-QUANTITY 0.82 0.83 0.82 105
B-TIME 0.69 0.69 0.69 212
B-WORK_OF_ART 0.71 0.72 0.71 166
I-CARDINAL 0.83 0.89 0.86 331
I-DATE 0.90 0.90 0.90 2011
I-EVENT 0.76 0.74 0.75 130
I-FAC 0.79 0.91 0.85 213
I-GPE 0.94 0.89 0.92 628
I-LAW 0.82 0.66 0.73 106
I-LOC 0.89 0.83 0.86 180
I-MONEY 0.94 0.96 0.95 685
I-NORP 0.98 0.91 0.94 160
I-ORDINAL 0.00 0.00 0.00 4
I-ORG 0.92 0.93 0.93 2406
I-PERCENT 0.96 0.95 0.96 523
I-PERSON 0.97 0.94 0.96 1412
I-PRODUCT 0.81 0.81 0.81 69
I-QUANTITY 0.87 0.92 0.89 206
I-TIME 0.68 0.73 0.70 255
I-WORK_OF_ART 0.72 0.66 0.69 337
O 0.99 0.99 0.99 131815
accuracy 0.98 152728
macro avg 0.83 0.82 0.82 152728
weighted avg 0.98 0.98 0.98 152728
processed 152728 tokens with 11257 phrases; found: 11394 phrases; correct: 10001.
accuracy: 90.30%; (non-O)
accuracy: 98.10%; precision: 87.77%; recall: 88.84%; FB1: 88.31
CARDINAL: precision: 83.37%; recall: 84.17%; FB1: 83.77 944
DATE: precision: 83.84%; recall: 86.14%; FB1: 84.98 1646
EVENT: precision: 64.06%; recall: 65.08%; FB1: 64.57 64
FAC: precision: 69.38%; recall: 82.22%; FB1: 75.25 160
GPE: precision: 96.64%; recall: 91.25%; FB1: 93.87 2115
LANGUAGE: precision: 78.95%; recall: 68.18%; FB1: 73.17 19
LAW: precision: 54.76%; recall: 57.50%; FB1: 56.10 42
LOC: precision: 70.10%; recall: 79.89%; FB1: 74.67 204
MONEY: precision: 87.70%; recall: 88.54%; FB1: 88.11 317
NORP: precision: 93.60%; recall: 95.72%; FB1: 94.65 860
ORDINAL: precision: 82.41%; recall: 91.28%; FB1: 86.62 216
ORG: precision: 87.26%; recall: 89.30%; FB1: 88.27 1837
PERCENT: precision: 89.43%; recall: 89.68%; FB1: 89.56 350
PERSON: precision: 93.70%; recall: 95.02%; FB1: 94.36 2016
PRODUCT: precision: 68.29%; recall: 73.68%; FB1: 70.89 82
QUANTITY: precision: 78.57%; recall: 83.81%; FB1: 81.11 112
TIME: precision: 58.85%; recall: 62.74%; FB1: 60.73 226
WORK_OF_ART: precision: 61.96%; recall: 68.67%; FB1: 65.14 184