Description
NER model that detects professions and occupations in Spanish texts. Trained with the embeddings_scielowiki_300d embeddings, and the same WordEmbeddingsModel is needed in the pipeline.
Predicted Entities
ACTIVIDAD, PROFESION, SITUACION_LABORAL
Live Demo Open in Colab Copy S3 URI
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence = SentenceDetector() \
.setInputCols("document") \
.setOutputCol("sentence")
tokenizer = Tokenizer() \
.setInputCols("sentence") \
.setOutputCol("token")
word_embeddings = WordEmbeddingsModel.pretrained("embeddings_scielowiki_300d", "es", "clinical/models")\
.setInputCols(["document", "token"])\
.setOutputCol("embeddings")
clinical_ner = MedicalNerModel.pretrained("meddroprof_scielowiki", "es", "clinical/models")\
.setInputCols(["sentence", "token", "embeddings"])\
.setOutputCol("ner")
ner_converter = NerConverter() \
.setInputCols(["sentence", "token", "ner"]) \
.setOutputCol("ner_chunk")
pipeline = Pipeline(stages=[
document_assembler,
sentence,
tokenizer,
word_embeddings,
clinical_ner,
ner_converter])
sample_text = """La paciente es la mayor de 2 hermanos, tiene un hermano de 13 años estudiando 1o ESO. Sus padres son ambos ATS , trabajan en diferentes centros de salud estudiando 1o ESO"""
df = spark.createDataFrame([[sample_text]]).toDF("text")
result = pipeline.fit(df).transform(df)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentence = new SentenceDetector()
.setInputCols("document")
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols("sentence")
.setOutputCol("token")
val word_embeddings = WordEmbeddingsModel.pretrained("embeddings_scielowiki_300d", "es", "clinical/models")
.setInputCols(Array("document", "token"))
.setOutputCol("word_embeddings")
val clinical_ner = MedicalNerModel.pretrained("meddroprof_scielowiki", "es", "clinical/models")
.setInputCols(Array("sentence", "token", "embeddings"))
.setOutputCol("ner")
val ner_converter = new NerConverter()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("ner_chunk")
val pipeline = new Pipeline().setStages(Array(document_assembler,
sentence,
tokenizer,
word_embeddings,
clinical_ner,
ner_converter))
val data = Seq("""La paciente es la mayor de 2 hermanos, tiene un hermano de 13 años estudiando 1o ESO. Sus padres son ambos ATS , trabajan en diferentes centros de salud estudiando 1o ESO""").toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("es.med_ner.scielowiki").predict("""La paciente es la mayor de 2 hermanos, tiene un hermano de 13 años estudiando 1o ESO. Sus padres son ambos ATS , trabajan en diferentes centros de salud estudiando 1o ESO""")
Results
+---------------------------------------+-----------------+
|chunk |ner_label |
+---------------------------------------+-----------------+
|estudiando 1o ESO |SITUACION_LABORAL|
|ATS |PROFESION |
|trabajan en diferentes centros de salud|PROFESION |
|estudiando 1o ESO |SITUACION_LABORAL|
+---------------------------------------+-----------------+
Model Information
| Model Name: | meddroprof_scielowiki |
| Compatibility: | Healthcare NLP 4.2.2+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | es |
| Size: | 14.8 MB |
References
The model was trained with the MEDDOPROF data set:
The MEDDOPROF corpus is a collection of 1844 clinical cases from over 20 different specialties annotated with professions and employment statuses. The corpus was annotated by a team composed of linguists and clinical experts following specially prepared annotation guidelines, after several cycles of quality control and annotation consistency analysis before annotating the entire dataset. Figure 1 shows a screenshot of a sample manual annotation generated using the brat annotation tool.
Reference:
@article{meddoprof,
title={NLP applied to occupational health: MEDDOPROF shared task at IberLEF 2022 on automatic recognition, classification and normalization of professions and occupations from medical texts},
author={Lima-López, Salvador and Farré-Maduell, Eulàlia and Miranda-Escalada, Antonio and Brivá-Iglesias, Vicent and Krallinger, Martin},
journal = {Procesamiento del Lenguaje Natural},
volume = {67},
year={2022}
}
Benchmarking
label precision recall f1-score support
B-ACTIVIDAD 0.82 0.36 0.50 25
B-PROFESION 0.87 0.75 0.81 634
B-SITUACION_LABORAL 0.79 0.67 0.72 310
I-ACTIVIDAD 0.86 0.43 0.57 58
I-PROFESION 0.87 0.80 0.83 944
I-SITUACION_LABORAL 0.74 0.71 0.73 407
O 1.00 1.00 1.00 139880
accuracy - - 0.99 142258
macro-avg 0.85 0.67 0.74 142258
weighted-avg 0.99 0.99 0.99 142258