Description
The legclf_trust_agreement model is a Legal Longformer Document Classifier to classify if the document belongs to the class trust-agreement or not (Binary Classification).
Longformers have a restriction on 4096 tokens, so only the first 4096 tokens will be taken into account. We have realised that for the big majority of the documents in legal corpora, if they are clean and only contain the legal document without any extra information before, 4096 is enough to perform Document Classification.
If not, let us know and we can carry out another approach for you: getting chunks of 4096 tokens and average the embeddings, training with the averaged version, what means all document will be taken into account. But this theoretically should not be required.
Predicted Entities
trust-agreement, other
How to use
document_assembler = nlp.DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")\
tokenizer = nlp.Tokenizer()\
.setInputCols(["document"])\
.setOutputCol("token")
embeddings = nlp.LongformerEmbeddings.pretrained("legal_longformer_base", "en")\
.setInputCols("document", "token")\
.setOutputCol("embeddings")
sentence_embeddings = nlp.SentenceEmbeddings()\
.setInputCols(["document", "embeddings"])\
.setOutputCol("sentence_embeddings")\
.setPoolingStrategy("AVERAGE")
doc_classifier = legal.ClassifierDLModel.pretrained("legclf_trust_agreement", "en", "legal/models")\
.setInputCols(["sentence_embeddings"])\
.setOutputCol("category")
nlpPipeline = nlp.Pipeline(stages=[
document_assembler,
tokenizer,
embeddings,
sentence_embeddings,
doc_classifier])
df = spark.createDataFrame([["YOUR TEXT HERE"]]).toDF("text")
model = nlpPipeline.fit(df)
result = model.transform(df)
Results
+-------+
|result|
+-------+
|[trust-agreement]|
|[other]|
|[other]|
|[trust-agreement]|
Model Information
| Model Name: | legclf_trust_agreement |
| Compatibility: | Legal NLP 1.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence_embeddings] |
| Output Labels: | [class] |
| Language: | en |
| Size: | 21.8 MB |
References
Legal documents, scrapped from the Internet, and classified in-house + SEC documents
Benchmarking
label precision recall f1-score support
other 0.99 0.97 0.98 90
trust-agreement 0.93 0.97 0.95 38
accuracy - - 0.97 128
macro-avg 0.96 0.97 0.96 128
weighted-avg 0.97 0.97 0.97 128