Description
IMPORTANT: Don’t run this model on the whole legal agreement. Instead:
- Split by paragraphs. You can use notebook 1 in Finance or Legal as inspiration;
- Use the
legclf_cuad_wheras_clauseText Classifier to select only these paragraphs;
This is a Relation Extraction model to infer relations between elements in WHEREAS clauses, more specifically the SUBJECT, the ACTION and the OBJECT. There are two relations possible: has_subject and has_object. You can also use legpipe_whereas which includes this model and its NER and also depedency parsing, to carry out chunk extraction using grammatical features (the dependency tree). This model requires legner_whereas as an NER in the pipeline. It’s a md model with Unidirectional Relations, meaning that the model retrieves in chunk1 the left side of the relation (source), and in chunk2 the right side (target).
Predicted Entities
has_subject, has_object
How to use
documentAssembler = nlp.DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
tokenizer = nlp.Tokenizer()\
.setInputCols("document")\
.setOutputCol("token")
embeddings = nlp.RoBertaEmbeddings.pretrained("roberta_embeddings_legal_roberta_base","en") \
.setInputCols(["document", "token"]) \
.setOutputCol("embeddings")
ner_model = legal.NerModel.pretrained('legner_whereas', 'en', 'legal/models')\
.setInputCols(["document", "token", "embeddings"])\
.setOutputCol("ner")
ner_converter = nlp.NerConverter()\
.setInputCols(["document","token","ner"])\
.setOutputCol("ner_chunk")
reDL = legal.RelationExtractionDLModel\
.pretrained("legre_whereas_md", "en", "legal/models")\
.setPredictionThreshold(0.9)\
.setInputCols(["ner_chunk", "document"])\
.setOutputCol("relations")
pipeline = nlp.Pipeline(stages=[
documentAssembler,
tokenizer,
embeddings,
ner_model,
ner_converter,
reDL
])
empty_df = spark.createDataFrame([[""]]).toDF("text")
model = pipeline.fit(empty_df)
text = """
Central Expressway, Suite 200, Dallas, TX 75080.
Background
The Supplier wishes to appoint the Distributor as its non-exclusive distributor for the promotion and sale of the Products within the Territory (both as defined below), and the Distributor wishes to promote and sell the Products within the Territory on the terms of this agreement.
Agreed terms
1. """
data = spark.createDataFrame([[text]]).toDF("text")
model = pipeline.fit(data)
res = model.transform(data)
Results
+-----------+---------------+-------------+-----------+------------------------------------------------+---------------+-------------+-----------+------------------------------------------------+----------+
|relation |entity1 |entity1_begin|entity1_end|chunk1 |entity2 |entity2_begin|entity2_end|chunk2 |confidence|
+-----------+---------------+-------------+-----------+------------------------------------------------+---------------+-------------+-----------+------------------------------------------------+----------+
|has_subject|WHEREAS_ACTION |76 |92 |wishes to appoint |WHEREAS_SUBJECT|63 |74 |The Supplier |0.9994367 |
|has_subject|WHEREAS_OBJECT |94 |141 |the Distributor as its non-exclusive distributor|WHEREAS_SUBJECT|63 |74 |The Supplier |0.92683166|
|has_subject|WHEREAS_SUBJECT|236 |250 |the Distributor |WHEREAS_SUBJECT|63 |74 |The Supplier |0.9829159 |
|has_object |WHEREAS_ACTION |76 |92 |wishes to appoint |WHEREAS_OBJECT |94 |141 |the Distributor as its non-exclusive distributor|0.900727 |
|has_object |WHEREAS_OBJECT |94 |141 |the Distributor as its non-exclusive distributor|WHEREAS_OBJECT |279 |290 |the Products |0.96618503|
|has_subject|WHEREAS_ACTION |252 |268 |wishes to promote |WHEREAS_SUBJECT|236 |250 |the Distributor |0.99969923|
+-----------+---------------+-------------+-----------+------------------------------------------------+---------------+-------------+-----------+------------------------------------------------+----------+
Model Information
| Model Name: | legre_whereas_md |
| Compatibility: | Legal NLP 1.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Language: | en |
| Size: | 402.2 MB |
References
Manual annotations on CUAD dataset
Benchmarking
label Recall Precision F1 Support
has_object 0.974 0.991 0.983 116
has_subject 0.977 0.986 0.981 213
other 0.993 0.978 0.985 271
Avg 0.981 0.985 0.983 -
Weighted-Avg 0.983 0.983 0.983 -