Description
This is a NER model, aimed to be run only after detecting the DISPUTE_RESOL clause with a proper classifier (use legmulticlf_mnda_sections_paragraph_other for that purpose). It will extract the following entities: COURT_NAME, LAW_LOCATION , and RESOLUT_MEANS .
Predicted Entities
COURT_NAME, LAW_LOCATION, RESOLUT_MEANS
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_dispute_resolution", "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 = ["""In case no settlement can be reached through consultation within thirty ( 30 ) days after such dispute is raised, each party can submit such matter to China International Economic and Trade Arbitration Commission ( the "CIETAC") in accordance with its rules."""]
result = model.transform(spark.createDataFrame([text]).toDF("text"))
Results
+-------------------------------------------------------+-------------+
|chunk |ner_label |
+-------------------------------------------------------+-------------+
|consultation |RESOLUT_MEANS|
|China |LAW_LOCATION |
|International Economic and Trade Arbitration Commission|COURT_NAME |
+-------------------------------------------------------+-------------+
Model Information
| Model Name: | legner_nda_dispute_resolution |
| 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
COURT_NAME 0.86 0.89 0.88 75
LAW_LOCATION 0.79 0.85 0.82 79
RESOLUT_MEANS 0.88 0.88 0.88 58
micro-avg 0.84 0.87 0.85 212
macro-avg 0.84 0.87 0.86 212
weighted-avg 0.84 0.87 0.85 212