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
Biomarker, Biomarker_Result
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence_detector = SentenceDetector()\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
labels = ['Biomarker', 'Biomarker_Result'] # You can change the entities
pretrained_zero_shot_ner = PretrainedZeroShotNER().pretrained("zeroshot_ner_oncology_biomarker_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([["""The results of immunohistochemical examination showed that she tested negative for CK7, synaptophysin (Syn), chromogranin A (CgA), Muc5AC, human epidermal growth factor receptor-2 (HER2), and Muc6; positive for CK20, Muc1, Muc2, E-cadherin, and p53; the Ki-67 index was about 87% ."""]]).toDF("text")
result = pipeline.fit(data).transform(data)
document_assembler = nlp.DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence_detector = nlp.SentenceDetector()\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = nlp.Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
labels = ['Biomarker', 'Biomarker_Result'] # You can change the entities
pretrained_zero_shot_ner = medical.PretrainedZeroShotNER().pretrained("zeroshot_ner_oncology_biomarker_large", "en", "clinical/models")\
.setInputCols("sentence", "token")\
.setOutputCol("ner")\
.setPredictionThreshold(0.5)\
.setLabels(labels)
ner_converter = medical.NerConverterInternal()\
.setInputCols("sentence", "token", "ner")\
.setOutputCol("ner_chunk")
pipeline = nlp.Pipeline().setStages([
document_assembler,
sentence_detector,
tokenizer,
pretrained_zero_shot_ner,
ner_converter
])
data = spark.createDataFrame([["""The results of immunohistochemical examination showed that she tested negative for CK7, synaptophysin (Syn), chromogranin A (CgA), Muc5AC, human epidermal growth factor receptor-2 (HER2), and Muc6; positive for CK20, Muc1, Muc2, E-cadherin, and p53; the Ki-67 index was about 87% ."""]]).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 = ["Biomarker", "Biomarker_Resul"] # You can change the entities
val pretrained_zero_shot_ner = PretrainedZeroShotNER().pretrained("zeroshot_ner_oncology_biomarker_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("""The results of immunohistochemical examination showed that she tested negative for CK7, synaptophysin (Syn), chromogranin A (CgA), Muc5AC, human epidermal growth factor receptor-2 (HER2), and Muc6; positive for CK20, Muc1, Muc2, E-cadherin, and p53; the Ki-67 index was about 87% .""").toDF("text")
val result = pipeline.fit(data).transform(data)
Results
+----------------------------------------+-----+---+----------------+----------+
|chunk |begin|end|ner_label |confidence|
+----------------------------------------+-----+---+----------------+----------+
|negative |71 |78 |Biomarker_Result|0.9885888 |
|CK7 |84 |86 |Biomarker |0.9961201 |
|synaptophysin |89 |101|Biomarker |0.9968708 |
|Syn |104 |106|Biomarker |0.7719608 |
|chromogranin A |110 |123|Biomarker |0.9940532 |
|Muc5AC |132 |137|Biomarker |0.99384034|
|human epidermal growth factor receptor-2|140 |179|Biomarker |0.9571454 |
|HER2 |182 |185|Biomarker |0.8589217 |
|Muc6 |193 |196|Biomarker |0.9970079 |
|positive |199 |206|Biomarker_Result|0.9965706 |
|CK20 |212 |215|Biomarker |0.9980994 |
|Muc1 |218 |221|Biomarker |0.9856312 |
|Muc2 |224 |227|Biomarker |0.9621688 |
|E-cadherin |230 |239|Biomarker |0.98347497|
|p53 |246 |248|Biomarker |0.99129874|
|Ki-67 index |255 |265|Biomarker |0.7583356 |
|87% |277 |279|Biomarker_Result|0.94230235|
+----------------------------------------+-----+---+----------------+----------+
Model Information
| Model Name: | zeroshot_ner_oncology_biomarker_large |
| Compatibility: | Healthcare NLP 5.5.1+ |
| License: | Licensed |
| Edition: | Official |
| Language: | en |
| Size: | 1.6 GB |
Benchmarking
label precision recall f1-score support
Biomarker 0.9056 0.9210 0.9133 2178
Biomarker_Result 0.8310 0.8969 0.8627 1464
accuracy - - 0.9436 13697
macro avg 0.9022 0.9244 0.9128 13697
weighted avg 0.9449 0.9436 0.9441 13697