Edit on GitHub
  • John Snow Labs
  • Aug 11, 2022

Extract Negation and Uncertainty Entities from Spanish Medical Texts (BertForTokenClassification)

  • es
  • clinical
  • licensed
  • token_classification
  • bert
  • ner
  • negation
  • uncertainty
  • linguistics

Description

This Named Entity Recognition model is intended for detecting relevant entities from Spanish medical texts and trained using the BertForTokenClassification method from the transformers library and BERT based embeddings. The model detects Negation Trigger (NEG), Negation Scope (NSCO), Uncertainty Trigger (UNC) and Uncertainty Scope (USCO).

Predicted Entities

NEG, NSCO, UNC, USCO

Copy S3 URI

How to use

documentAssembler = DocumentAssembler()\
  .setInputCol("text")\
  .setOutputCol("document")

sentenceDetector = SentenceDetectorDLModel.pretrained()\
  .setInputCols(["document"])\
  .setOutputCol("sentence")

tokenizer = Tokenizer()\
  .setInputCols("sentence")\
  .setOutputCol("token")

tokenClassifier = MedicalBertForTokenClassifier.pretrained("bert_token_classifier_negation_uncertainty", "es", "clinical/models")\
  .setInputCols("token", "sentence")\
  .setOutputCol("label")\
  .setCaseSensitive(True)

ner_converter = NerConverter()\
  .setInputCols(["sentence","token","label"])\
  .setOutputCol("ner_chunk")

pipeline =  Pipeline(stages=[
                      documentAssembler,
                      sentenceDetector,
                      tokenizer,
                      tokenClassifier,
                      ner_converter])

model = pipeline.fit(spark.createDataFrame([[""]]).toDF("text"))

data = spark.createDataFrame(["Con diagnóstico probable de cirrosis hepática (no conocida previamente) y peritonitis espontanea primaria con tratamiento durante 8 dias con ceftriaxona en el primer ingreso (no se realizó paracentesis control por escasez de liquido). Lesión tumoral en hélix izquierdo de 0,5 cms. de diámetro susceptible de ca basocelular perlado."], StringType()).toDF("text")
                              
result = model.transform(data)

val documentAssembler = new DocumentAssembler()
  .setInputCol("text")
  .setOutputCol("document")

val sentenceDetector = SentenceDetectorDLModel.pretrained()
  .setInputCols("document")
  .setOutputCol("sentence")

val tokenizer = new Tokenizer()
  .setInputCols("sentence")
  .setOutputCol("token")

val tokenClassifier = MedicalBertForTokenClassifier.pretrained("bert_token_classifier_negation_uncertainty", "es", "clinical/models")
  .setInputCols(Array("token", "sentence"))
  .setOutputCol("label")
  .setCaseSensitive(True)

val ner_converter = new NerConverter()
  .setInputCols(Array("sentence","token","label"))
  .setOutputCol("ner_chunk")

val pipeline =  new Pipeline().setStages(Array(
                      documentAssembler,
                      sentenceDetector,
                      tokenizer,
                      tokenClassifier,
                      ner_converter))

val data = Seq(Array("""Con diagnóstico probable de cirrosis hepática (no conocida previamente) y peritonitis espontanea primaria con tratamiento durante 8 dias con ceftriaxona en el primer ingreso (no se realizó paracentesis control por escasez de liquido). Lesión tumoral en hélix izquierdo de 0,5 cms. de diámetro susceptible de ca basocelular perlado.""")).toDS().toDF("text")

val result = pipeline.fit(data).transform(data)

import nlu
nlu.load("es.classify.bert_token.negation_uncertainty").predict("""Con diagnóstico probable de cirrosis hepática (no conocida previamente) y peritonitis espontanea primaria con tratamiento durante 8 dias con ceftriaxona en el primer ingreso (no se realizó paracentesis control por escasez de liquido). Lesión tumoral en hélix izquierdo de 0,5 cms. de diámetro susceptible de ca basocelular perlado.""")

Results

+------------------------------------------------------+---------+
|chunk                                                 |ner_label|
+------------------------------------------------------+---------+
|probable                                              |UNC      |
|de cirrosis hepática                                  |USCO     |
|no                                                    |NEG      |
|conocida previamente                                  |NSCO     |
|no                                                    |NEG      |
|se realizó paracentesis control por escasez de liquido|NSCO     |
|susceptible de                                        |UNC      |
|ca basocelular perlado                                |USCO     |
+------------------------------------------------------+---------+

Model Information

Model Name: bert_token_classifier_negation_uncertainty
Compatibility: Healthcare NLP 4.0.2+
License: Licensed
Edition: Official
Input Labels: [sentence, token]
Output Labels: [ner]
Language: es
Size: 410.0 MB
Case sensitive: true
Max sentence length: 512

References

The model is prepared using the reference paper: “NLP applied to occupational health: MEDDOPROF shared task at IberLEF 2021 on automatic recognition, classification and normalization of professions and occupations from medical texts”, Salvador Lima-Lopez, Eulalia Farr ́e-Maduell, Antonio Miranda-Escalada, Vicent Briva-Iglesias and Martin Krallinger. Procesamiento del Lenguaje Natural, Revista nº 67, septiembre de 2021, pp. 243-256.

Benchmarking

       label  precision    recall  f1-score   support
       B-NEG     0.9599    0.9667    0.9633      1833
       I-NEG     0.9216    0.9276    0.9246       152
       B-UNC     0.9040    0.8898    0.8968       508
       I-UNC     0.8772    0.8242    0.8499       182
      B-USCO     0.9164    0.8983    0.9073       708
      I-USCO     0.8473    0.8596    0.8534      2350
      B-NSCO     0.9475    0.9560    0.9517      2022
      I-NSCO     0.9323    0.9345    0.9334      5774
   micro-avg     0.9207    0.9239    0.9223     13529
   macro-avg     0.9133    0.9071    0.9101     13529
weighted-avg     0.9208    0.9239    0.9223     13529