Description
This is an NER model aimed to be used in notice clauses, to retrieve entities as NOTICE_METHOD, NOTICE_PARTY, ADDRESS, EMAIL, etc. Make sure you run this model only on notice clauses, after you filter them using legclf_notice_clause
Predicted Entities
ADDRESS, DEPARTMENT, EMAIL, FAX, NAME, NOTICE_METHOD, NOTICE_PARTY, PERSON, PHONE, TITLE
How to use
document_assembler = 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_notice_clause", "en", "legal/models") \
.setInputCols(["document", "token", "embeddings"]) \
.setOutputCol("ner")
ner_converter = nlp.NerConverter() \
.setInputCols(["document","token","ner"]) \
.setOutputCol("ner_chunk")
pipeline = nlp.Pipeline(stages=[
document_assembler,
tokenizer,
embeddings,
ner_model,
ner_converter
])
empty_df = spark.createDataFrame([['']]).toDF("text")
ner_model = pipeline.fit(empty_df)
data = spark.createDataFrame([["""Source: FUELCELL ENERGY INC, 8-K, 11/6/2019
ExxonMobil: ExxonMobil Research and Engineering Company 1545 Route 22 East Annandale, NJ 08801-0900 Attention: Timothy Barckholtz, Senior Scientific Advisor Email: tim.barckholtz@exxonmobil.com FCE: FuelCell Energy, Inc. 782"""]]).toDF("text")
result = ner_model.transform(data)
Results
+---------------------------------------------------------------------+------------+
|ner_chunk |label |
+---------------------------------------------------------------------+------------+
|ExxonMobil |NOTICE_PARTY|
|ExxonMobil Research and Engineering Company |NAME |
|1545 Route 22 East Annandale, NJ 08801-0900 |ADDRESS |
|Timothy Barckholtz |PERSON |
|Senior Scientific Advisor |TITLE |
|tim.barckholtz@exxonmobil.com |EMAIL |
+---------------------------------------------------------------------+------------+
Model Information
| Model Name: | legner_notice_clause |
| Compatibility: | Legal NLP 1.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | en |
| Size: | 1.1 MB |
References
In-house dataset
Benchmarking
label precision recall f1-score support
ADDRESS 0.86 0.94 0.90 141
DEPARTMENT 0.75 0.27 0.40 11
EMAIL 0.92 1.00 0.96 48
FAX 0.65 0.88 0.75 51
NAME 0.78 0.79 0.79 140
NOTICE_METHOD 0.74 0.80 0.77 353
NOTICE_PARTY 0.77 0.85 0.81 103
PERSON 0.91 0.94 0.92 114
PHONE 0.60 0.47 0.53 19
TITLE 0.76 0.90 0.82 80
micro-avg 0.78 0.85 0.81 1060
macro-avg 0.77 0.79 0.77 1060
weighted-avg 0.79 0.85 0.81 1060