Description
Named Entity Recognition annotators allow for a generic model to be trained by using a Deep Learning architecture (Char CNNs - BiLSTM - CRF - word embeddings) inspired on a former state of the art model for NER: Chiu & Nicols, Named Entity Recognition with Bidirectional LSTM,CNN.
Deidentification NER (Portuguese) is a Named Entity Recognition model that annotates text to find protected health information that may need to be de-identified. It detects 15 entities.
This NER model is trained with a combination of custom datasets with several data augmentation mechanisms.
Predicted Entities
PATIENT, HOSPITAL, DATE, ORGANIZATION, CITY, ID, STREET, SEX, EMAIL, ZIP, PROFESSION, PHONE, COUNTRY, DOCTOR, AGE
Live Demo Open in Colab Copy S3 URI
How to use
documentAssembler = nlp.DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentenceDetector = nlp.SentenceDetectorDLModel.pretrained("sentence_detector_dl","xx")\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = nlp.Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
embeddings = nlp.WordEmbeddingsModel.pretrained("w2v_cc_300d", "pt")\
.setInputCols(["sentence","token"])\
.setOutputCol("word_embeddings")
clinical_ner = medical.NerModel.pretrained("ner_deid_subentity", "pt", "clinical/models")\
.setInputCols(["sentence","token","word_embeddings"])\
.setOutputCol("ner")
ner_converter = nlp.NerConverter()\
.setInputCols(["sentence","token","ner"])\
.setOutputCol("ner_chunk")
nlpPipeline = nlp.Pipeline(stages=[
documentAssembler,
sentenceDetector,
tokenizer,
embeddings,
clinical_ner,
ner_converter])
text = ['''
Detalhes do paciente.
Nome do paciente: Pedro Gonçalves
NHC: 2569870.
Endereço: Rua Das Flores 23.
Cidade/ Província: Porto.
Código Postal: 21754-987.
Dados de cuidados.
Data de nascimento: 10/10/1963.
Idade: 53 anos Sexo: Homen
Data de admissão: 17/06/2016.
Doutora: Maria Santos
''']
data = spark.createDataFrame([text]).toDF("text")
result = nlpPipeline.fit(data).transform(data)
val documentAssembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentenceDetector = SentenceDetectorDLModel.pretrained("sentence_detector_dl","xx")
.setInputCols(Array("document"))
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols(Array("sentence"))
.setOutputCol("token")
embeddings = WordEmbeddingsModel.pretrained("w2v_cc_300d", "pt")
.setInputCols(Array("sentence","token"))
.setOutputCol("word_embeddings")
clinical_ner = MedicalNerModel.pretrained("ner_deid_subentity", "pt", "clinical/models")
.setInputCols(Array("sentence","token","word_embeddings"))
.setOutputCol("ner")
val ner_converter = new NerConverter()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("ner_chunk")
val pipeline = new Pipeline().setStages(Array(
documentAssembler,
sentenceDetector,
tokenizer,
embeddings,
clinical_ner,
ner_converter))
val text = """Detalhes do paciente.
Nome do paciente: Pedro Gonçalves
NHC: 2569870.
Endereço: Rua Das Flores 23.
Cidade/ Província: Porto.
Código Postal: 21754-987.
Dados de cuidados.
Data de nascimento: 10/10/1963.
Idade: 53 anos Sexo: Homen
Data de admissão: 17/06/2016.
Doutora: Maria Santos"""
val data = Seq(text).toDF("text")
val results = pipeline.fit(data).transform(data)
import nlu
nlu.load("pt.med_ner.deid.subentity").predict("""
Detalhes do paciente.
Nome do paciente: Pedro Gonçalves
NHC: 2569870.
Endereço: Rua Das Flores 23.
Cidade/ Província: Porto.
Código Postal: 21754-987.
Dados de cuidados.
Data de nascimento: 10/10/1963.
Idade: 53 anos Sexo: Homen
Data de admissão: 17/06/2016.
Doutora: Maria Santos
""")
Results
+-----------------+------------+
|chunk |ner_label |
+-----------------+------------+
|Pedro Gonçalves |PATIENT |
|2569870 |ID |
|Rua Das Flores 23|STREET |
|Porto |ORGANIZATION|
|21754-987 |ID |
|10/10/1963 |DATE |
|53 |AGE |
|17/06/2016 |DATE |
|Maria Santos |DOCTOR |
+-----------------+------------+
Model Information
| Model Name: | ner_deid_subentity |
| Compatibility: | Healthcare NLP 3.4.2+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | pt |
| Size: | 15.0 MB |
References
- Custom John Snow Labs datasets
- Data augmentation techniques
Benchmarking
label tp fp fn total precision recall f1
PATIENT 2142.0 186.0 59.0 2201.0 0.9201 0.9732 0.9459
HOSPITAL 248.0 10.0 46.0 294.0 0.9612 0.8435 0.8986
DATE 1306.0 26.0 15.0 1321.0 0.9805 0.9886 0.9845
ORGANIZATION 3038.0 31.0 156.0 3194.0 0.9899 0.9512 0.9701
CITY 1836.0 58.0 15.0 1851.0 0.9694 0.9919 0.9805
ID 56.0 8.0 7.0 63.0 0.875 0.8889 0.8819
STREET 155.0 0.0 0.0 155.0 1.0 1.0 1.0
SEX 658.0 20.0 19.0 677.0 0.9705 0.9719 0.9712
EMAIL 131.0 0.0 0.0 131.0 1.0 1.0 1.0
ZIP 125.0 2.0 0.0 125.0 0.9843 1.0 0.9921
PROFESSION 237.0 15.0 39.0 276.0 0.9405 0.8587 0.8977
PHONE 64.0 2.0 0.0 64.0 0.9697 1.0 0.9846
COUNTRY 502.0 27.0 74.0 576.0 0.949 0.8715 0.9086
DOCTOR 461.0 35.0 31.0 492.0 0.9294 0.937 0.9332
AGE 538.0 17.0 8.0 546.0 0.9694 0.9853 0.9773
macro - - - - - - 0.9551
micro - - - - - - 0.9619