Description
This is a single entity model that generalises all posology concepts into one and finds longest available chunks of drugs. It is trained using embeddings_clinical so please use the same embeddings in the pipeline.
Predicted Entities
DRUG.
How to use
Use as part of an nlp pipeline with the following stages: DocumentAssembler, SentenceDetector, Tokenizer, WordEmbeddingsModel, NerDLModel. Add the NerConverter to the end of the pipeline to convert entity tokens into full entity chunks.
...
word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")
clinical_ner = NerDLModel.pretrained("ner_drugs_greedy", "en", "clinical/models") \
.setInputCols(["sentence", "token", "embeddings"]) \
.setOutputCol("ner")
...
nlp_pipeline = Pipeline(stages=[document_assembler, sentence_detector, tokenizer, word_embeddings, clinical_ner, ner_converter])
model = nlp_pipeline.fit(spark.createDataFrame([[""]]).toDF("text"))
results = model.transform(spark.createDataFrame([["DOSAGE AND ADMINISTRATION The initial dosage of hydrocortisone tablets may vary from 20 mg to 240 mg of hydrocortisone per day depending on the specific disease entity being treated."]]).toDF("text"))
...
val word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
val clinical_ner = NerDLModel.pretrained("ner_drugs_greedy", "en", "clinical/models")
.setInputCols(Array("sentence", "token", "embeddings"))
.setOutputCol("ner")
...
val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, tokenizer, word_embeddings, clinical_ner, ner_converter))
val result = pipeline.fit(Seq.empty["DOSAGE AND ADMINISTRATION The initial dosage of hydrocortisone tablets may vary from 20 mg to 240 mg of hydrocortisone per day depending on the specific disease entity being treated."].toDS.toDF("text")).transform(data)
Results
+-----------------------------------+------------+
| chunk | ner_label |
+-----------------------------------+------------+
| hydrocortisone tablets | DRUG |
| 20 mg to 240 mg of hydrocortisone | DRUG |
+-----------------------------------+------------+
Model Information
| Model Name: | ner_drugs_greedy |
| Type: | ner |
| Compatibility: | Spark NLP 2.6.5+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | en |
| Dependencies: | embeddings_clinical |
Data Source
Trained on augmented i2b2_med7 + FDA dataset with embeddings_clinical, https://www.i2b2.org/NLP/Medication.
Benchmarking
label tp fp fn prec rec f1
I-DRUG 37858 4166 3338 0.90086615 0.91897273 0.9098294
B-DRUG 29926 2006 1756 0.937179 0.9445742 0.9408621
tp: 67784 fp: 6172 fn: 5094 labels: 2
Macro-average prec: 0.91902256, rec: 0.9317734, f1: 0.92535406
Micro-average prec: 0.916545, rec: 0.93010235, f1: 0.9232739