Description
This relation extraction model links Biomarker and Oncogene extractions to their corresponding Biomarker_Result extractions.
Predicted Entities
is_finding_of, O
Live Demo Open in Colab Copy S3 URI
How to use
Use relation pairs to include only the combinations of entities that are relevant in your case.
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl_healthcare","en","clinical/models")\
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer() \
.setInputCols(["sentence"]) \
.setOutputCol("token")
word_embeddings = WordEmbeddingsModel().pretrained("embeddings_clinical", "en", "clinical/models")\
.setInputCols(["sentence", "token"]) \
.setOutputCol("embeddings")
ner = MedicalNerModel.pretrained("ner_oncology_wip", "en", "clinical/models") \
.setInputCols(["sentence", "token", "embeddings"]) \
.setOutputCol("ner")
ner_converter = NerConverterInternal() \
.setInputCols(["sentence", "token", "ner"]) \
.setOutputCol("ner_chunk")
pos_tagger = PerceptronModel.pretrained("pos_clinical", "en", "clinical/models") \
.setInputCols(["sentence", "token"]) \
.setOutputCol("pos_tags")
dependency_parser = DependencyParserModel.pretrained("dependency_conllu", "en") \
.setInputCols(["sentence", "pos_tags", "token"]) \
.setOutputCol("dependencies")
re_ner_chunk_filter = RENerChunksFilter()\
.setInputCols(["ner_chunk", "dependencies"])\
.setOutputCol("re_ner_chunk")\
.setMaxSyntacticDistance(10)\
.setRelationPairs(["Biomarker-Biomarker_Result", "Biomarker_Result-Biomarker", "Oncogene-Biomarker_Result", "Biomarker_Result-Oncogene"])
re_model = RelationExtractionDLModel.pretrained("redl_oncology_biomarker_result_biobert_wip", "en", "clinical/models")\
.setInputCols(["re_ner_chunk", "sentence"])\
.setOutputCol("relation_extraction")
pipeline = Pipeline(stages=[document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
ner,
ner_converter,
pos_tagger,
dependency_parser,
re_ner_chunk_filter,
re_model])
data = spark.createDataFrame([["Immunohistochemistry was negative for thyroid transcription factor-1 and napsin A. The test was positive for ER and PR, and negative for HER2."]]).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_healthcare","en","clinical/models")
.setInputCols(Array("document"))
.setOutputCol("sentence")
val tokenizer = new Tokenizer()
.setInputCols(Array("sentence"))
.setOutputCol("token")
val word_embeddings = WordEmbeddingsModel().pretrained("embeddings_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("embeddings")
val ner = MedicalNerModel.pretrained("ner_oncology_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 pos_tagger = PerceptronModel.pretrained("pos_clinical", "en", "clinical/models")
.setInputCols(Array("sentence", "token"))
.setOutputCol("pos_tags")
val dependency_parser = DependencyParserModel.pretrained("dependency_conllu", "en")
.setInputCols(Array("sentence", "pos_tags", "token"))
.setOutputCol("dependencies")
val re_ner_chunk_filter = new RENerChunksFilter()
.setInputCols(Array("ner_chunk", "dependencies"))
.setOutputCol("re_ner_chunk")
.setMaxSyntacticDistance(10)
.setRelationPairs(Array("Biomarker-Biomarker_Result", "Biomarker_Result-Biomarker", "Oncogene-Biomarker_Result", "Biomarker_Result-Oncogene"))
val re_model = RelationExtractionDLModel.pretrained("redl_oncology_biomarker_result_biobert_wip", "en", "clinical/models")
.setInputCols(Array("re_ner_chunk", "sentence"))
.setOutputCol("relation_extraction")
val pipeline = new Pipeline().setStages(Array(document_assembler,
sentence_detector,
tokenizer,
word_embeddings,
ner,
ner_converter,
pos_tagger,
dependency_parser,
re_ner_chunk_filter,
re_model))
val data = Seq("Immunohistochemistry was negative for thyroid transcription factor-1 and napsin A. The test was positive for ER and PR, and negative for HER2.").toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("en.relation.oncology_biomarker_result_biobert_wip").predict("""Immunohistochemistry was negative for thyroid transcription factor-1 and napsin A. The test was positive for ER and PR, and negative for HER2.""")
Results
+-------------+----------------+-------------+-----------+--------+----------------+-------------+-----------+--------------------+----------+
| relation| entity1|entity1_begin|entity1_end| chunk1| entity2|entity2_begin|entity2_end| chunk2|confidence|
+-------------+----------------+-------------+-----------+--------+----------------+-------------+-----------+--------------------+----------+
|is_finding_of|Biomarker_Result| 25| 32|negative| Biomarker| 38| 67|thyroid transcrip...|0.99808085|
|is_finding_of|Biomarker_Result| 25| 32|negative| Biomarker| 73| 78| napsin|0.99637383|
|is_finding_of|Biomarker_Result| 96| 103|positive| Biomarker| 109| 110| ER|0.99221414|
|is_finding_of|Biomarker_Result| 96| 103|positive| Biomarker| 116| 117| PR| 0.9893672|
| O|Biomarker_Result| 96| 103|positive| Oncogene| 137| 140| HER2| 0.9986272|
| O| Biomarker| 109| 110| ER|Biomarker_Result| 124| 131| negative| 0.9999089|
| O| Biomarker| 116| 117| PR|Biomarker_Result| 124| 131| negative| 0.9998932|
|is_finding_of|Biomarker_Result| 124| 131|negative| Oncogene| 137| 140| HER2|0.98810333|
+-------------+----------------+-------------+-----------+--------+----------------+-------------+-----------+--------------------+----------+
Model Information
| Model Name: | redl_oncology_biomarker_result_biobert_wip |
| Compatibility: | Healthcare NLP 4.2.4+ |
| License: | Licensed |
| Edition: | Official |
| Language: | en |
| Size: | 401.7 MB |
References
In-house annotated oncology case reports.
Benchmarking
label recall precision f1
O 0.93 0.97 0.95
is_finding_of 0.97 0.93 0.95
macro-avg 0.95 0.95 0.95