Named Entity Recognition - OntoNotes RoBERTa (ner_ontonotes_roberta_base)

Description

ner_ontonotes_roberta_base is a Named Entity Recognition (or NER) model trained on OntoNotes 5.0. It can extract up to 18 entities such as people, places, organizations, money, time, date, etc.

This model uses the pretrained roberta_base model from the RoBertaEmbeddings annotator as an input.

Predicted Entities

CARDINAL, DATE, EVENT, FAC, GPE, LANGUAGE, LAW, LOC, MONEY, NORP, ORDINAL, ORG, PERCENT, PERSON, PRODUCT, QUANTITY, TIME, WORK_OF_ART

Live Demo Open in Colab Download

How to use

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

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

embeddings = RoBertaEmbeddings\
      .pretrained('roberta_base', 'en')\
      .setInputCols(["token", "document"])\
      .setOutputCol("embeddings")

ner_model = NerDLModel.pretrained('ner_ontonotes_roberta_base', 'en') \
    .setInputCols(['document', 'token', 'embeddings']) \
    .setOutputCol('ner')

ner_converter = NerConverter() \
    .setInputCols(['document', 'token', 'ner']) \
    .setOutputCol('entities')

pipeline = Pipeline(stages=[
    document_assembler, 
    tokenizer,
    embeddings,
    ner_model,
    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 embeddings = RoBertaEmbeddings.pretrained("roberta_base", "en")
    .setInputCols("document", "token") 
    .setOutputCol("embeddings")

val ner_model = NerDLModel.pretrained("ner_ontonotes_roberta_base", "en") 
    .setInputCols("document"', "token", "embeddings") 
    .setOutputCol("ner")

val ner_converter = NerConverter() 
    .setInputCols("document", "token", "ner") 
    .setOutputCol("entities")

val pipeline = new Pipeline().setStages(Array(document_assembler, tokenizer, embeddings, ner_model, ner_converter))

val example = Seq.empty["My name is John!"].toDS.toDF("text")

val result = pipeline.fit(example).transform(example)
import nlu

text = ["My name is John!"]

ner_df = nlu.load('en.ner.ner_ontonotes_roberta_base').predict(text, output_level='token')

Model Information

Model Name: ner_ontonotes_roberta_base
Type: ner
Compatibility: Spark NLP 3.2.0+
License: Open Source
Edition: Official
Input Labels: [sentence, token, embeddings]
Output Labels: [ner]
Language: en

Data Source

https://catalog.ldc.upenn.edu/LDC2013T19

Benchmarking

             precision    recall  f1-score   support

   B-CARDINAL       0.85      0.87      0.86       935
       B-DATE       0.86      0.89      0.88      1602
      B-EVENT       0.65      0.51      0.57        63
        B-FAC       0.79      0.56      0.66       135
        B-GPE       0.97      0.92      0.94      2240
   B-LANGUAGE       0.82      0.41      0.55        22
        B-LAW       0.59      0.57      0.58        40
        B-LOC       0.81      0.74      0.77       179
      B-MONEY       0.89      0.91      0.90       314
       B-NORP       0.92      0.95      0.93       841
    B-ORDINAL       0.81      0.88      0.84       195
        B-ORG       0.86      0.92      0.89      1795
    B-PERCENT       0.93      0.92      0.93       349
     B-PERSON       0.93      0.93      0.93      1988
    B-PRODUCT       0.72      0.63      0.67        76
   B-QUANTITY       0.81      0.81      0.81       105
       B-TIME       0.64      0.67      0.65       212
B-WORK_OF_ART       0.72      0.57      0.64       166
   I-CARDINAL       0.84      0.81      0.82       331
       I-DATE       0.87      0.92      0.90      2011
      I-EVENT       0.71      0.66      0.69       130
        I-FAC       0.79      0.67      0.72       213
        I-GPE       0.91      0.89      0.90       628
        I-LAW       0.77      0.64      0.70       106
        I-LOC       0.86      0.71      0.78       180
      I-MONEY       0.94      0.97      0.95       685
       I-NORP       0.91      0.85      0.88       160
    I-ORDINAL       0.00      0.00      0.00         4
        I-ORG       0.89      0.94      0.91      2406
    I-PERCENT       0.95      0.96      0.96       523
     I-PERSON       0.94      0.93      0.94      1412
    I-PRODUCT       0.70      0.70      0.70        69
   I-QUANTITY       0.89      0.90      0.89       206
       I-TIME       0.64      0.74      0.68       255
I-WORK_OF_ART       0.75      0.55      0.64       337
            O       0.99      0.99      0.99    131815

     accuracy                           0.98    152728
    macro avg       0.80      0.76      0.78    152728
 weighted avg       0.98      0.98      0.98    152728


processed 152728 tokens with 11257 phrases; found: 11277 phrases; correct: 9868.
accuracy:  89.00%; (non-O)
accuracy:  97.82%; precision:  87.51%; recall:  87.66%; FB1:  87.58
         CARDINAL: precision:  84.05%; recall:  86.20%; FB1:  85.11  959
             DATE: precision:  84.18%; recall:  87.02%; FB1:  85.57  1656
            EVENT: precision:  63.27%; recall:  49.21%; FB1:  55.36  49
              FAC: precision:  77.08%; recall:  54.81%; FB1:  64.07  96
              GPE: precision:  96.03%; recall:  91.74%; FB1:  93.84  2140
         LANGUAGE: precision:  81.82%; recall:  40.91%; FB1:  54.55  11
              LAW: precision:  56.41%; recall:  55.00%; FB1:  55.70  39
              LOC: precision:  79.63%; recall:  72.07%; FB1:  75.66  162
            MONEY: precision:  87.81%; recall:  89.49%; FB1:  88.64  320
             NORP: precision:  91.10%; recall:  93.70%; FB1:  92.38  865
          ORDINAL: precision:  80.66%; recall:  87.69%; FB1:  84.03  212
              ORG: precision:  84.05%; recall:  89.25%; FB1:  86.57  1906
          PERCENT: precision:  90.20%; recall:  89.68%; FB1:  89.94  347
           PERSON: precision:  92.66%; recall:  92.66%; FB1:  92.66  1988
          PRODUCT: precision:  68.66%; recall:  60.53%; FB1:  64.34  67
         QUANTITY: precision:  80.95%; recall:  80.95%; FB1:  80.95  105
             TIME: precision:  60.00%; recall:  63.68%; FB1:  61.78  225
      WORK_OF_ART: precision:  65.38%; recall:  51.20%; FB1:  57.43  130