Datasets:

pie

/

argmicro

like 0

This adjusts the converter for TextDocumentWithLabeledSpansAndBinaryRelations to produce data comparable with the literature, e.g. End-to-end Argument Mining with Cross-corpora Multi-task Learning. However, this irreversibly modifies and simplifies the data, i.e. the original data can not be re-constructed.

This should live behind a parameter.