Description
This model extracts demographic information related to Social Determinants of Health from various kinds of biomedical documents.
Predicted Entities
Family_Member, Age, Gender, Geographic_Entity, Race_Ethnicity, Language, Spiritual_Beliefs
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl", "en")\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
clinical_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")
ner_model = MedicalNerModel.pretrained("ner_sdoh_demographics_wip", "en", "clinical/models")\
.setInputCols(["sentence", "token", "embeddings"])\
.setOutputCol("ner")
ner_converter = NerConverterInternal()\
.setInputCols(["sentence", "token", "ner"])\
.setOutputCol("ner_chunk")
pipeline = Pipeline(stages=[
document_assembler,
sentence_detector,
tokenizer,
clinical_embeddings,
ner_model,
ner_converter
])
sample_texts = ["SOCIAL HISTORY: He is a former tailor from Korea.",
"He lives alone,single and no children.",
"Pt is a 61 years old married, Caucasian, Catholic woman. Pt speaks English reasonably well."]
data = spark.createDataFrame(sample_texts, StringType()).toDF("text")
result = pipeline.fit(data).transform(data)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl", "en")
.setInputCols("document")
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols("sentence")
.setOutputCol("token")
val clinical_embeddings = WordEmbeddingsModel.pretrained("embeddings_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
val ner_model = MedicalNerModel.pretrained("ner_sdoh_demographics_wip", "en", "clinical/models")
.setInputCols(Array("sentence", "token", "embeddings"))
.setOutputCol("ner")
val ner_converter = new NerConverterInternal()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("ner_chunk")
val pipeline = new Pipeline().setStages(Array(
document_assembler,
sentence_detector,
tokenizer,
clinical_embeddings,
ner_model,
ner_converter
))
val data = Seq("Pt is a 61 years old married, Caucasian, Catholic woman. Pt speaks English reasonably well.").toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
Results
+-----------------+-----+---+------------+
|ner_label |begin|end|chunk |
+-----------------+-----+---+------------+
|Gender |16 |17 |He |
|Geographic_Entity|43 |47 |Korea |
|Gender |0 |1 |He |
|Family_Member |29 |36 |children |
|Age |8 |19 |61 years old|
|Race_Ethnicity |30 |38 |Caucasian |
|Spiritual_Beliefs|41 |48 |Catholic |
|Gender |50 |54 |woman |
|Language |67 |73 |English |
+-----------------+-----+---+------------+
Model Information
| Model Name: | ner_sdoh_demographics_wip |
| Compatibility: | Healthcare NLP 5.3.3+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | en |
| Size: | 852.5 KB |
Benchmarking
label tp fp fn total precision recall f1
Age 1346.0 73.0 74.0 1420.0 0.948555 0.947887 0.948221
Spiritual_Beliefs 100.0 13.0 16.0 116.0 0.884956 0.862069 0.873362
Family_Member 4468.0 134.0 43.0 4511.0 0.970882 0.990468 0.980577
Race_Ethnicity 56.0 0.0 13.0 69.0 1.000000 0.811594 0.896000
Gender 9825.0 67.0 247.0 10072.0 0.993227 0.975477 0.984272
Geographic_Entity 225.0 9.0 29.0 254.0 0.961538 0.885827 0.922131
Language 51.0 9.0 5.0 56.0 0.850000 0.910714 0.879310