src.dackar.RCA.ner.hybrid_ner.postprocess

Classes

Postprocessor

Postprocessing stage:

Module Contents

class src.dackar.RCA.ner.hybrid_ner.postprocess.Postprocessor[source]
Postprocessing stage:
  • flatten accepted ResolvedSpan entities

  • apply canonicalization (optional later)

  • filter by confidence thresholds (optional later)

v0.1: just flatten accepted spans.

apply(doc_id, decisions, relations, schema)[source]
Parameters:
Return type:

src.dackar.RCA.ner.hybrid_ner.models.PipelineResult

_prune_nested_entities(ents)[source]
Parameters:

ents (List[src.dackar.RCA.ner.hybrid_ner.models.ResolvedSpan])

Return type:

List[src.dackar.RCA.ner.hybrid_ner.models.ResolvedSpan]

_should_drop_nested(inner, outer)[source]

Drop only when the nested span is likely redundant, not when it is a useful atomic entity for RCA.

Parameters:
Return type:

bool