Description
Deep learning named entity recognition model for assertions. The SparkNLP deep learning model (NerDL) is inspired by a former state of the art model for NER: Chiu & Nicols, Named Entity Recognition with Bidirectional LSTM-CNN.
Predicted Entities
hypothetical, present, absent, possible, conditional, associated_with_someone_else.
How to use
Use as part of an nlp pipeline with the following stages: DocumentAssembler, SentenceDetector, Tokenizer, WordEmbeddingsModel, NerDLModel, NerConverter, AssertionDLModel.
documentAssembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentenceDetector = SentenceDetector()\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")
clinical_ner = MedicalNerModel.pretrained("ner_clinical", "en", "clinical/models") \
.setInputCols(["sentence", "token", "embeddings"]) \
.setOutputCol("ner")
ner_converter = NerConverter() \
.setInputCols(["sentence", "token", "ner"]) \
.setOutputCol("ner_chunk")
clinical_assertion = AssertionDLModel.pretrained("assertion_dl_large", "en", "clinical/models") \
.setInputCols(["sentence", "ner_chunk", "embeddings"]) \
.setOutputCol("assertion")
nlpPipeline = Pipeline(stages=[documentAssembler, sentenceDetector, tokenizer, word_embeddings, clinical_ner, ner_converter, clinical_assertion])
model = nlpPipeline.fit(spark.createDataFrame([["Patient with severe fever and sore throat. He shows no stomach pain and he maintained on an epidural and PCA for pain control. He also became short of breath with climbing a flight of stairs. After CT, lung tumor located at the right lower lobe. Father with Alzheimer."]]).toDF("text"))
light_model = LightPipeline(model)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentence_detector = new SentenceDetector()
.setInputCols("document")
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols("sentence")
.setOutputCol("token")
val word_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
val clinical_ner = MedicalNerModel.pretrained("ner_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token", "embeddings"))
.setOutputCol("ner")
val nerConverter = new NerConverter()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("ner_chunk")
val clinical_assertion = AssertionDLModel.pretrained("assertion_dl_large", "en", "clinical/models")
.setInputCols(Array("sentence", "ner_chunk", "embeddings"))
.setOutputCol("assertion")
val pipeline = new Pipeline().setStages(Array(documentAssembler, sentenceDetector, tokenizer, word_embeddings, clinical_ner, nerConverter, clinical_assertion))
val data = Seq("""Patient with severe fever and sore throat. He shows no stomach pain and he maintained on an epidural and PCA for pain control. He also became short of breath with climbing a flight of stairs. After CT, lung tumor located at the right lower lobe. Father with Alzheimer.""").toDS().toDF("text")
val result = pipeline.fit(data).transform(data)
Results
The output is a dataframe with a sentence per row and an "assertion" column containing all of the assertion labels in the sentence. The assertion column also contains assertion character indices, and other metadata. To get only the entity chunks and assertion labels, without the metadata, select "ner_chunk.result" and "assertion.result" from your output dataframe.
chunks entities assertion
0 severe fever PROBLEM present
1 sore throat PROBLEM present
2 stomach pain PROBLEM absent
3 an epidural TREATMENT present
4 PCA TREATMENT present
5 pain control PROBLEM present
6 short of breath PROBLEM conditional
7 CT TEST present
8 lung tumor PROBLEM present
9 Alzheimer PROBLEM associated_with_someone_else
Model Information
| Model Name: | assertion_dl_large |
| Type: | ner |
| Compatibility: | Spark NLP 2.5.0+ |
| Edition: | Official |
| License: | Licensed |
| Input Labels: | [sentence, ner_chunk, embeddings] |
| Output Labels: | [assertion] |
| Language: | [en] |
| Case sensitive: | false |
Data Source
Trained with augmented version of 2010 i2b2/VA dataset on concepts, assertions, and relations in clinical text with ‘embeddings_clinical’. https://portal.dbmi.hms.harvard.edu/projects/n2c2-nlp/
Benchmarking
prec rec f1
absent 0.97 0.91 0.94
associated_with_someone_else 0.93 0.87 0.90
conditional 0.70 0.33 0.44
hypothetical 0.91 0.82 0.86
possible 0.81 0.59 0.68
present 0.93 0.98 0.95
micro avg 0.93 0.93 0.93
macro avg 0.87 0.75 0.80