XLM-RoBERTa Token Classification Base - NER CoNLL (xlm_roberta_base_token_classifier_conll03)

Description

XLM-RoBERTa 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.

xlm_roberta_base_token_classifier_conll03 is a fine-tuned XLM-RoBERTa 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 TFXLMRobertaForTokenClassification to train this model and used XlmRoBertaForTokenClassification annotator in Spark NLP 🚀 for prediction at scale!

Predicted Entities

Download Copy S3 URI

How to use

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

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

tokenClassifier = XlmRoBertaForTokenClassification \
      .pretrained('xlm_roberta_base_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 = XlmRoBertaForTokenClassification.pretrained("xlm_roberta_base_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)
import nlu
nlu.load("en.ner.xlm_roberta.conll.base").predict("""My name is John!""")

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: xlm_roberta_base_token_classifier_conll03
Compatibility: Spark NLP 3.3.0+
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

            precision    recall  f1-score   support

       B-LOC       0.95      0.84      0.89      1837
      B-MISC       0.87      0.86      0.86       922
       B-ORG       0.82      0.91      0.86      1341
       B-PER       0.96      0.96      0.96      1842
       I-LOC       0.89      0.84      0.86       257
      I-MISC       0.84      0.76      0.80       346
       I-ORG       0.88      0.91      0.90       751
       I-PER       0.98      0.97      0.97      1307
           O       0.99      1.00      1.00     42759

    accuracy                           0.98     51362
   macro avg       0.91      0.89      0.90     51362
weighted avg       0.98      0.98      0.98     51362



processed 51362 tokens with 5942 phrases; found: 5900 phrases; correct: 5257.
accuracy:  90.17%; (non-O)
accuracy:  98.08%; precision:  89.10%; recall:  88.47%; FB1:  88.79
              LOC: precision:  94.22%; recall:  83.34%; FB1:  88.45  1625
             MISC: precision:  84.40%; recall:  83.30%; FB1:  83.84  910
              ORG: precision:  79.60%; recall:  89.34%; FB1:  84.19  1505
              PER: precision:  94.62%; recall:  95.55%; FB1:  95.08  1860