Legal Question Answering (RoBerta, CUAD, Large)

documentAssembler = nlp.MultiDocumentAssembler()\
        .setInputCols(["question", "context"])\
        .setOutputCols(["document_question", "document_context"])

spanClassifier = nlp.RoBertaForQuestionAnswering.pretrained("legqa_roberta_cuad_large","en", "legal/models") \
.setInputCols(["document_question", "document_context"]) \
.setOutputCol("answer") \
.setCaseSensitive(True)

pipeline = nlp.Pipeline().setStages([
documentAssembler,
spanClassifier
])

text = """THIS CREDIT AGREEMENT is dated as of April 29, 2010, and is made by and
      among P.H. GLATFELTER COMPANY, a Pennsylvania corporation ( the "COMPANY") and
      certain of its subsidiaries. Identified on the signature pages hereto (each a
      "BORROWER" and collectively, the "BORROWERS"), each of the GUARANTORS (as
      hereinafter defined), the LENDERS (as hereinafter defined), PNC BANK, NATIONAL
      ASSOCIATION, in its capacity as agent for the Lenders under this Agreement
      (hereinafter referred to in such capacity as the "ADMINISTRATIVE AGENT"), and,
      for the limited purpose of public identification in trade tables, PNC CAPITAL
      MARKETS LLC and CITIZENS BANK OF PENNSYLVANIA, as joint arrangers and joint
      bookrunners, and CITIZENS BANK OF PENNSYLVANIA, as syndication agent.""".replace('\n',' ')
        
        
question = ['"Highlight the parts (if any) of this contract related to "Parties" that should be reviewed by a lawyer. Details: The two or more parties who signed the contract"']

qt = [ [q,text] for q in questions    ]

example = spark.createDataFrame(qt).toDF("question", "context")

result = pipeline.fit(example).transform(example)

result.select('document_question.result', 'answer.result').show(truncate=False)