Version 1.19.0 of the library comes with a topic classification multi-label model using the INSTRUCTOR embedding.
New Multilabel Model on Legal Stack Exchange topics with INSTRUCTOR
The model can identify more than thirty topics, listed below:
`business`, `california`, `canada`, `civil-law`, `constitutional-law`, `consumer-protection`, `contract-law`, `copyright`, `corporate-law`, `criminal-law`, `employment`, `england-and-wales`, `european-union`, `fraud`, `gdpr`, `germany`, `intellectual-property`, `international`, `internet`, `landlord`, `legal-terms`, `liability`, `licensing`, `police`, `privacy`, `property`, `real-estate`, `rental-property`, `software`, `tax-law`, `terms-of-service`, `trademark`, `traffic`, `united-kingdom`, `united-states`, `us-constitution`
Please note that some categories from the original data were removed due to limited examples. The model depends on the INSTRUCTOR embedding, which can be used in a Spark NLP pipeline as follows:
embeddings = ( nlp.InstructorEmbeddings.pretrained("instructor_large", "en") .setInstruction("Represent for multilabel classification:") .setInputCols(["document"]) .setOutputCol("sentence_embeddings") ) classifierdl = ( nlp.MultiClassifierDLModel.pretrained( "legmulticlf_law_stack_exchange", "en", "legal/models" ) .setInputCols(["sentence_embeddings"]) .setOutputCol("class") ) clf_pipeline = nlp.Pipeline( stages=[document_assembler, embeddings, classifierdl] )
For the given example:
At one point, I saw this label on Coke cans and wondered: is this legally enforceable? If it’s not, is it possible for a retailer in any way to disallow the resale of an item purchased? Something like, I don’t know, maybe a license you must agree to to purchase said item? This comes in the larger context of these new tech releases (GPUs, consoles) and how the producers/retailers could legally prevent scalpers.
We can extract the topics with the pipeline (text formatted for visualization purposes):
df = spark.createDataFrame( [ [ """I've seen this label on Coke cans at one point, and I was wondering: is this legally enforceable? If it's not, is it possible for a retailer in any way to disallow the resale of an item purchased? Something like, I don't know, maybe a license you have to agree to in order to be allowed to purchase said item? This comes in the larger context of these new tech releases (GPUs, consoles) and how the producers/retailers could legally prevent scalpers.""" ] ] ).toDF("text") model = clf_pipeline.fit(df) result = model.transform(df) result.select("class.result").show(truncate=False)
Obtaining the following results:
+------------------------+ result | +------------------------+ |[contract-law, business]| +------------------------+
The topics obtained were
We have 30-day free licenses for you with technical support from our legal team of experts and SME. This trial includes access to more than 926 models, including Classification, NER, Relation Extraction, Similarity Search, Summarization, Sentiment Analysis, Question Answering, etc., and 120+ legal language models.
Just go to https://www.johnsnowlabs.com/install/ and follow the instructions!
How to run
Legal NLP (NLP for contracts) is extremely easy to run on both clusters and driver-only environments using
pip install johnsnowlabs
from johnsnowlabs import nlp nlp.install(force_browser=True)
# Start Spark Session spark = nlp.start()
# Import the Legal NLP module from johnsnowlabs import legal
For alternative installation methods of how to install in specific environments, please check the docs.