Edit on GitHub
  • John Snow Labs
  • Apr 10, 2023

Legal NER for NDA (Non-solicitation Clauses)

  • en
  • legal
  • licensed
  • ner
  • nda
  • non_solicitation

Description

This is a NER model, aimed to be run only after detecting the NON_SOLIC clause with a proper classifier (use legmulticlf_mnda_sections_paragraph_other model for that purpose). It will extract the following entities: NON_SOLIC_ACTION, NON_SOLIC_OBJECT, NON_SOLIC_IND_OBJECT, and NON_SOLIC_SUBJECT.

Predicted Entities

NON_SOLIC_ACTION, NON_SOLIC_OBJECT, NON_SOLIC_IND_OBJECT, NON_SOLIC_SUBJECT

Download Copy S3 URI

How to use

document_assembler = nlp.DocumentAssembler()\
        .setInputCol("text")\
        .setOutputCol("document")
        
sentence_detector = nlp.SentenceDetector()\
        .setInputCols(["document"])\
        .setOutputCol("sentence")

tokenizer = nlp.Tokenizer()\
        .setInputCols(["sentence"])\
        .setOutputCol("token")

embeddings = nlp.RoBertaEmbeddings.pretrained("roberta_embeddings_legal_roberta_base","en") \
        .setInputCols(["sentence", "token"]) \
        .setOutputCol("embeddings")\
        .setMaxSentenceLength(512)\
        .setCaseSensitive(True)

ner_model = legal.NerModel.pretrained("legner_nda_non_solicitation", "en", "legal/models")\
        .setInputCols(["sentence", "token", "embeddings"])\
        .setOutputCol("ner")

ner_converter = nlp.NerConverter()\
        .setInputCols(["sentence", "token", "ner"])\
        .setOutputCol("ner_chunk")

nlpPipeline = nlp.Pipeline(stages=[
        document_assembler,
        sentence_detector,
        tokenizer,
        embeddings,
        ner_model,
        ner_converter
])

empty_data = spark.createDataFrame([[""]]).toDF("text")

model = nlpPipeline.fit(empty_data)

text = ["""During the 12-month period commencing on the date of this Agreement, each Party agrees that it will not permit any of its officers, directors, or employees or any direct or indirect subsidiary or any employment agency retained by a Party or any such subsidiary, in each case who is or becomes aware of the negotiation of a possible Transaction between the Parties, to solicit for employment with such Party or with any of its direct or indirect subsidiaries any employee of the other Party or any of its direct or indirect subsidiaries;"""]

result = model.transform(spark.createDataFrame([text]).toDF("text"))

Results

+----------+--------------------+
|chunk     |ner_label           |
+----------+--------------------+
|Party     |NON_SOLIC_SUBJECT   |
|officers  |NON_SOLIC_IND_OBJECT|
|directors |NON_SOLIC_IND_OBJECT|
|employees |NON_SOLIC_IND_OBJECT|
|agency    |NON_SOLIC_SUBJECT   |
|solicit   |NON_SOLIC_ACTION    |
|employment|NON_SOLIC_OBJECT    |
|employee  |NON_SOLIC_IND_OBJECT|
+----------+--------------------+

Model Information

Model Name: legner_nda_non_solicitation
Compatibility: Legal NLP 1.0.0+
License: Licensed
Edition: Official
Input Labels: [sentence, token, embeddings]
Output Labels: [ner]
Language: en
Size: 16.3 MB

References

In-house annotations on the Non-disclosure Agreements

Benchmarking

label                 precision  recall  f1-score  support 
NON_SOLIC_ACTION      1.00       0.95    0.98      22      
NON_SOLIC_IND_OBJECT  0.67       0.90    0.76      29      
NON_SOLIC_OBJECT      0.81       0.94    0.87      18      
NON_SOLIC_SUBJECT     0.86       0.86    0.86      21      
micro-avg             0.80       0.91    0.85      90      
macro-avg             0.83       0.91    0.87      90      
weighted-avg          0.82       0.91    0.86      90