BERT Token Classification Large - NER CoNLL (bert_large_token_classifier_conll03)

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_conll03 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: location (LOC), organizations (ORG), person (PER), and Miscellaneous (MISC).

We used TFBertForTokenClassification to train this model and used BertForTokenClassification annotator in Spark NLP 🚀 for prediction at scale!

Predicted Entities

PER, LOC, ORG, MISC

Download

How to use

document_assembler = DocumentAssembler() \
    .setInputCol('text') \
    .setOutputCol('document')

tokenizer = Tokenizer() \
    .setInputCols(['document']) \
    .setOutputCol('token')

tokenClassifier = BertForTokenClassification \
      .pretrained('bert_large_token_classifier_conll03', '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_conll03", "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)

Results

+------------------------------------------------------------------------------------+
 |result                                                                              |
 +------------------------------------------------------------------------------------+
 |[B-PER, I-PER, O, O, O, B-LOC, O, O, O, B-LOC, O, O, O, O, B-PER, O, O, O, O, B-LOC]|
 +------------------------------------------------------------------------------------+

Model Information

Model Name: bert_large_token_classifier_conll03
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://www.clips.uantwerpen.be/conll2003/ner/

Benchmarking

Dev:

         precision    recall  f1-score   support

       B-LOC       0.96      0.90      0.93      1837
       I-ORG       0.93      0.95      0.94       751
      I-MISC       0.91      0.87      0.89       346
       I-LOC       0.91      0.94      0.93       257
       I-PER       0.99      0.98      0.99      1307
      B-MISC       0.94      0.90      0.92       922
       B-ORG       0.88      0.95      0.91      1341
       B-PER       0.98      0.98      0.98      1842

   micro avg       0.95      0.94      0.94      8603
   macro avg       0.94      0.93      0.93      8603
weighted avg       0.95      0.94      0.94      8603



processed 51362 tokens with 5942 phrases; found: 5915 phrases; correct: 5497.
accuracy:  93.99%; (non-O)
accuracy:  98.80%; precision:  92.93%; recall:  92.51%; FB1:  92.72
              LOC: precision:  95.25%; recall:  89.49%; FB1:  92.28  1726
             MISC: precision:  90.35%; recall:  88.39%; FB1:  89.36  902
              ORG: precision:  86.75%; recall:  93.21%; FB1:  89.86  1441
              PER: precision:  96.86%; recall:  97.07%; FB1:  96.96  1846


Test:

         precision    recall  f1-score   support

       B-LOC       0.93      0.89      0.91      1668
       I-ORG       0.86      0.94      0.90       835
      I-MISC       0.68      0.75      0.71       216
       I-LOC       0.87      0.86      0.87       257
       I-PER       0.98      0.98      0.98      1156
      B-MISC       0.84      0.82      0.83       702
       B-ORG       0.87      0.92      0.90      1661
       B-PER       0.97      0.96      0.97      1617

   micro avg       0.91      0.92      0.91      8112
   macro avg       0.88      0.89      0.88      8112
weighted avg       0.91      0.92      0.92      8112



processed 46435 tokens with 5648 phrases; found: 5682 phrases; correct: 5104.
accuracy:  92.01%; (non-O)
accuracy:  98.09%; precision:  89.83%; recall:  90.37%; FB1:  90.10
              LOC: precision:  92.39%; recall:  88.85%; FB1:  90.59  1604
             MISC: precision:  79.75%; recall:  80.20%; FB1:  79.97  706
              ORG: precision:  85.50%; recall:  91.27%; FB1:  88.29  1773
              PER: precision:  96.50%; recall:  95.42%; FB1:  95.96  1599