Relation extraction between body parts and direction entities (ReDL).

Description

Relation extraction between body parts entities like Internal_organ_or_component, External_body_part_or_region etc. and Direction entities like upper, lower in clinical texts. 1 : Shows the body part and direction entity are related, 0 : Shows the body part and direction entity are not related.

Predicted Entities

0, 1

Open in Colab Download

How to use

...
words_embedder = WordEmbeddingsModel() \
    .pretrained("embeddings_clinical", "en", "clinical/models") \
    .setInputCols(["sentences", "tokens"]) \
    .setOutputCol("embeddings")
ner_tagger = NerDLModel() \
    .pretrained("jsl_ner_wip_greedy_clinical", "en", "clinical/models") \
    .setInputCols(["sentences", "tokens", "embeddings"]) \
    .setOutputCol("ner_tags")
ner_converter = NerConverter() \
    .setInputCols(["sentences", "tokens", "ner_tags"]) \
    .setOutputCol("ner_chunks")
dependency_parser = DependencyParserModel() \
    .pretrained("dependency_conllu", "en") \
    .setInputCols(["sentences", "pos_tags", "tokens"]) \
    .setOutputCol("dependencies")

# Set a filter on pairs of named entities which will be treated as relation candidates
re_ner_chunk_filter = RENerChunksFilter() \
    .setInputCols(["ner_chunks", "dependencies"])\
    .setMaxSyntacticDistance(10)\
    .setOutputCol("re_ner_chunks").setRelationPairs(['direction-internal_organ_or_component', 'internal_organ_or_component-direction'])

# The dataset this model is trained to is sentence-wise. 
# This model can also be trained on document-level relations - in which case, while predicting, use "document" instead of "sentence" as input.
re_model = RelationExtractionDLModel()\
    .pretrained('redl_bodypart_direction_biobert', 'en', "clinical/models") \
    .setPredictionThreshold(0.5)\
    .setInputCols(["re_ner_chunks", "sentences"]) \
    .setOutputCol("relations")

pipeline = Pipeline(stages=[documenter, sentencer, tokenizer, pos_tagger, words_embedder, ner_tagger, ner_converter, dependency_parser, re_ner_chunk_filter, re_model])

text ="MRI demonstrated infarction in the upper brain stem , left cerebellum and  right basil ganglia"
p_model = pipeline.fit(spark.createDataFrame([[text]]).toDF("text"))
result = p_model.transform(data)
...
val words_embedder = WordEmbeddingsModel()
    .pretrained("embeddings_clinical", "en", "clinical/models")
    .setInputCols(Array("sentences", "tokens"))
    .setOutputCol("embeddings")
val ner_tagger = NerDLModel()
    .pretrained("ner_clinical", "en", "clinical/models")
    .setInputCols(Array("sentences", "tokens", "embeddings"))
    .setOutputCol("ner_tags")
val ner_converter = NerConverter()
    .setInputCols(Array("sentences", "tokens", "ner_tags"))
    .setOutputCol("ner_chunks")
val dependency_parser = DependencyParserModel()
    .pretrained("dependency_conllu", "en")
    .setInputCols(Array("sentences", "pos_tags", "tokens"))
    .setOutputCol("dependencies")

// Set a filter on pairs of named entities which will be treated as relation candidates
val re_ner_chunk_filter = RENerChunksFilter()
    .setInputCols(Array("ner_chunks", "dependencies"))
    .setMaxSyntacticDistance(10)
    .setOutputCol("re_ner_chunks").setRelationPairs(Array("direction-internal_organ_or_component", "internal_organ_or_component-direction"))

// The dataset this model is trained to is sentence-wise. 
// This model can also be trained on document-level relations - in which case, while predicting, use "document" instead of "sentence" as input.
val re_model = RelationExtractionDLModel()
    .pretrained("redl_bodypart_direction_biobert", "en", "clinical/models")
    .setPredictionThreshold(0.5)
    .setInputCols(Array("re_ner_chunks", "sentences"))
    .setOutputCol("relations")
val pipeline = new Pipeline().setStages(Array(documenter, sentencer, tokenizer, pos_tagger, words_embedder, ner_tagger, ner_converter, dependency_parser, re_ner_chunk_filter, re_model))

val data = Seq("MRI demonstrated infarction in the upper brain stem , left cerebellum and  right basil ganglia").toDF("text")
val result = pipeline.fit(data).transform(data)

Results

| index | relations | entity1                     | entity1_begin | entity1_end | chunk1     | entity2                     | entity2_end | entity2_end | chunk2        | confidence |
|-------|-----------|-----------------------------|---------------|-------------|------------|-----------------------------|-------------|-------------|---------------|------------|
| 0     | 1         | Direction                   | 35            | 39          | upper      | Internal_organ_or_component | 41          | 50          | brain stem    | 0.9999989  |
| 1     | 0         | Direction                   | 35            | 39          | upper      | Internal_organ_or_component | 59          | 68          | cerebellum    | 0.99992585 |
| 2     | 0         | Direction                   | 35            | 39          | upper      | Internal_organ_or_component | 81          | 93          | basil ganglia | 0.9999999  |
| 3     | 0         | Internal_organ_or_component | 41            | 50          | brain stem | Direction                   | 54          | 57          | left          | 0.999811   |
| 4     | 0         | Internal_organ_or_component | 41            | 50          | brain stem | Direction                   | 75          | 79          | right         | 0.9998203  |
| 5     | 1         | Direction                   | 54            | 57          | left       | Internal_organ_or_component | 59          | 68          | cerebellum    | 1.0        |
| 6     | 0         | Direction                   | 54            | 57          | left       | Internal_organ_or_component | 81          | 93          | basil ganglia | 0.97616416 |
| 7     | 0         | Internal_organ_or_component | 59            | 68          | cerebellum | Direction                   | 75          | 79          | right         | 0.953046   |
| 8     | 1         | Direction                   | 75            | 79          | right      | Internal_organ_or_component | 81          | 93          | basil ganglia | 1.0        |

Model Information

Model Name: redl_bodypart_direction_biobert
Compatibility: Spark NLP for Healthcare 2.7.3+
License: Licensed
Edition: Official
Language: en

Data Source

Trained on an internal dataset.

Benchmarking

Relation           Recall Precision        F1   Support
0                   0.856     0.873     0.865       153
1                   0.986     0.984     0.985      1347
Avg.                0.921     0.929     0.925