task normalization for text matching

[pfliu-nlp]

Following tasks could be merged, potentially:

• ranking:

  DataLab/datalabs/tasks/ranking.py

  Line 10 in cf10dbd

  class Ranking(TaskTemplate):

• ArgumentPairIdentification:

  DataLab/datalabs/tasks/argument_pair_identification.py

  Line 10 in cf10dbd

  class ArgumentPairIdentification(TaskTemplate):

• qa-mutliple-choice-withoutcontext:

  DataLab/datalabs/tasks/question_answering.py

  Line 272 in cf10dbd

  class QuestionAnsweringMultipleChoiceWithoutContext(QuestionAnsweringMultipleChoice):

[pfliu-nlp]

@register_task(TaskType.argument_pair_identification)
@dataclass
class ArgumentPairIdentification(Ranking):
    task: TaskType = TaskType.argument_pair_identification
    context_column: str = "quotation"  
    utterance_column: str = "replies"  # a list of replies, including positive and negative reply
    label_column: str = "label"           

[pfliu-nlp]

"quotation": Value("string"), # context = quotation
"replies": Sequence(Value("string")), # utterance =
"label":Value("int32"), # label = label
"quotation_context": Value("string"),
"replies_context": Sequence(Value("string")),