Description
Zero-shot Named Entity Recognition (NER) enables the identification of entities in text with minimal effort. By leveraging pre-trained language models and contextual understanding, zero-shot NER extends entity recognition capabilities to new domains and languages. While the model card includes default labels as examples, it is important to highlight that users are not limited to these labels. The model is designed to support any set of entity labels, allowing users to adapt it to their specific use cases. For best results, it is recommended to use labels that are conceptually similar to the provided defaults.
Predicted Entities
DATE, PATIENT, COUNTRY, PROFESSION, AGE, CITY, STATE, DOCTOR, HOSPITAL, IDNUM, ORGANIZATION, PHONE, STREET, ZIP
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence_detector = SentenceDetector()\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
labels = ["AGE", "CITY", "COUNTRY", "DATE", "DOCTOR", "HOSPITAL", "IDNUM", "ORGANIZATION",
"PATIENT", "PHONE", "PROFESSION", "STATE", "STREET", "ZIP"]
pretrained_zero_shot_ner = PretrainedZeroShotNER().pretrained("zeroshot_ner_deid_subentity_docwise_large", "en", "clinical/models")\
.setInputCols("sentence", "token")\
.setOutputCol("ner")\
.setPredictionThreshold(0.5)\
.setLabels(labels)
ner_converter = NerConverterInternal()\
.setInputCols("sentence", "token", "ner")\
.setOutputCol("ner_chunk")
pipeline = Pipeline().setStages([
document_assembler,
sentence_detector,
tokenizer,
pretrained_zero_shot_ner,
ner_converter
])
data = spark.createDataFrame([["""Emily Davis, a 34-year-old woman, Dr. Michael Johnson cares wit her, at CarePlus Clinic, located at 456 Elm Street, NewYork, NY has recommended starting insulin therapy. She has an appointment scheduled for March 15, 2024."""]]).toDF("text")
result = pipeline.fit(data).transform(data)
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")
labels = Array("AGE", "CITY", "COUNTRY", "DATE", "DOCTOR", "HOSPITAL", "IDNUM", "ORGANIZATION",
"PATIENT", "PHONE", "PROFESSION", "STATE", "STREET", "ZIP")
val pretrained_zero_shot_ner = PretrainedZeroShotNER().pretrained("zeroshot_ner_deid_subentity_docwise_large", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("ner")
.setPredictionThreshold(0.5)
.setLabels(labels)
val ner_converter = new NerConverterInternal()
.setInputCols(Array("sentence", "token", "ner"))
.setOutputCol("ner_chunk")
val pipeline = new Pipeline().setStages(Array(
document_assembler,
sentence_detector,
tokenizer,
pretrained_zero_shot_ner,
ner_converter
))
val data = Seq("""Emily Davis, a 34-year-old woman, Dr. Michael Johnson cares wit her, at CarePlus Clinic, located at 456 Elm Street, NewYork, NY has recommended starting insulin therapy. She has an appointment scheduled for March 15, 2024.""").toDF("text")
val result = pipeline.fit(data).transform(data)
Results
+---------------+-----+---+---------+----------+
|chunk |begin|end|ner_label|confidence|
+---------------+-----+---+---------+----------+
|Emily Davis |1 |11 |PATIENT |0.9994733 |
|34-year-old |16 |26 |AGE |0.99977213|
|Michael Johnson|39 |53 |DOCTOR |0.9995535 |
|CarePlus Clinic|73 |87 |HOSPITAL |0.9834484 |
|456 Elm Street |101 |114|STREET |0.9990957 |
|NewYork |117 |123|CITY |0.9991258 |
|NY |126 |127|STATE |0.9988438 |
|March 15, 2024 |208 |221|DATE |0.999948 |
+---------------+-----+---+---------+----------+
Model Information
| Model Name: | zeroshot_ner_deid_subentity_docwise_large |
| Compatibility: | Healthcare NLP 5.5.1+ |
| License: | Licensed |
| Edition: | Official |
| Language: | en |
| Size: | 1.6 GB |
Benchmarking
label precision recall f1-score support
AGE 0.9392 0.9348 0.9370 1074
CITY 0.8367 0.9467 0.8883 525
COUNTRY 0.8750 0.8652 0.8701 178
DATE 0.9895 0.9811 0.9853 7995
DOCTOR 0.9766 0.9513 0.9638 5134
HOSPITAL 0.8680 0.9130 0.8899 2276
IDNUM 0.9322 0.8785 0.9046 955
ORGANIZATION 0.7696 0.7778 0.7737 189
PATIENT 0.8996 0.9590 0.9283 2364
PHONE 0.9375 0.9146 0.9259 492
PROFESSION 0.9213 0.9474 0.9341 494
STATE 0.8671 0.9503 0.9068 302
STREET 0.9882 0.9653 0.9766 605
ZIP 0.8874 0.9949 0.9381 198
accuracy 0.9946 337866
macro avg 0.9124 0.9318 0.9214 337866
weighted avg 0.9947 0.9946 0.9946 337866