Description
This model can generate SQL queries from natural questions. It is based on a small-size LLM, which is finetuned by John Snow Labs on a dataset having a schema with the same schema that MIMIC-III has.
Predicted Entities
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
text2sql = Text2SQL.pretrained("text2sql_mimicsql", "en", "clinical/models")\
.setInputCols(["document"])\
.setOutputCol("sql")
pipeline = Pipeline(stages=[
document_assembler,
text2sql
])
text = "Calulate the total number of patients who had icd9 code 5771"
data = spark.createDataFrame([[text]]).toDF("text")
pipeline = Pipeline(stages=[document_assembler, text2sql])
result= pipeline.fit(data).transform(data)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val text2sql = new Text2SQL.pretrained("text2sql_mimicsql", "en", "clinical/models")
.setInputCols(["document"])
.setOutputCol("sql")
val pipeline = new Pipeline().setStages(Array(document_assembler, text2sql ))
val text = """Calulate the total number of patients who had icd9 code 5771"""
val data = Seq(Array(text)).toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
Results
[
SELECT COUNT ( DISTINCT DEMOGRAPHIC."SUBJECT_ID" )
FROM DEMOGRAPHIC
INNER JOIN PROCEDURES on DEMOGRAPHIC.HADM_ID = PROCEDURES.HADM_ID
WHERE PROCEDURES."ICD9_CODE" = "5771"
]
Model Information
| Model Name: | text2sql_mimicsql |
| Compatibility: | Healthcare NLP 5.0.1+ |
| License: | Licensed |
| Edition: | Official |
| Language: | en |
| Size: | 3.0 GB |