Description
This model detects Drug, Dosage, and administration instructions in text using pretrained NER model.
Predicted Entities
Live Demo Open in Colab Download Copy S3 URI
How to use
documentAssembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentenceDetector = SentenceDetectorDLModel.pretrained("sentence_detector_dl_healthcare","en","clinical/models") \
.setInputCols(["document"]) \
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical_medium", "en", "clinical/models")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")
posology_ner = MedicalNerModel.pretrained('ner_posology_emb_clinical_medium' , "en", "clinical/models") \
.setInputCols(["sentence", "token", "embeddings"]) \
.setOutputCol("posology_ner")
posology_ner_converter = NerConverter() \
.setInputCols(["sentence", "token", "posology_ner"]) \
.setOutputCol("posology_ner_chunk")
posology_ner_pipeline = Pipeline(stages=[
documentAssembler,
sentenceDetector,
tokenizer,
word_embeddings,
posology_ner,
posology_ner_converter])
empty_data = spark.createDataFrame([[""]]).toDF("text")
posology_ner_model = posology_ner_pipeline.fit(empty_data)
results = posology_ner_model.transform(spark.createDataFrame([["The patient has been advised Aspirin 81 milligrams QDay. insulin 50 units in a.m. HCTZ 50 mg QDay. Nitroglycerin 1/150 sublingually."]]).toDF("text"))
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl_healthcare","en","clinical/models")
.setInputCols("document")
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols("sentence")
.setOutputCol("token")
val word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical_medium", "en", "clinical/models")\
.setInputCols(Array("sentence", "token"))\
.setOutputCol("embeddings")
val posology_ner_model = MedicalNerModel.pretrained('ner_posology_emb_clinical_large' "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("posology_ner")
val posology_ner_converter = new NerConverter()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("posology_ner_chunk")
val posology_pipeline = new PipelineModel().setStages(Array(document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
posology_ner_model,
posology_ner_converter))
val data = Seq(""" The patient has been advised Aspirin 81 milligrams QDay. insulin 50 units in a.m. HCTZ 50 mg QDay. Nitroglycerin 1/150 sublingually.""").toDS.toDF("text")
val result = model.fit(data).transform(data)
Results
| | chunks | begin | end | entities |
|---:|:--------------|--------:|------:|:-----------|
| 0 | Aspirin | 29 | 35 | DRUG |
| 1 | 81 milligrams | 37 | 49 | STRENGTH |
| 2 | QDay | 51 | 54 | FREQUENCY |
| 3 | insulin | 57 | 63 | DRUG |
| 4 | 50 units | 65 | 72 | STRENGTH |
| 5 | HCTZ | 82 | 85 | DRUG |
| 6 | 50 mg | 87 | 91 | STRENGTH |
| 7 | QDay | 93 | 96 | FREQUENCY |
| 8 | Nitroglycerin | 99 | 111 | DRUG |
| 9 | 1/150 | 113 | 117 | STRENGTH |
| 10 | sublingually | 119 | 130 | ROUTE |
Model Information
| Model Name: | ner_posology_emb_clinical_medium |
| Compatibility: | Healthcare NLP 4.3.2+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [document, token, embeddings] |
| Output Labels: | [ner] |
| Language: | en |
| Size: | 2.8 MB |
Benchmarking
precision recall f1-score support
DRUG 0.91 0.91 0.91 2252
STRENGTH 0.88 0.93 0.91 2290
FREQUENCY 0.90 0.94 0.92 1782
DURATION 0.78 0.84 0.81 463
DOSAGE 0.66 0.63 0.65 476
ROUTE 0.89 0.89 0.89 394
FORM 0.86 0.76 0.81 773
micro avg 0.87 0.89 0.88 8430
macro avg 0.84 0.84 0.84 8430
weighted avg 0.87 0.89 0.88 8430