src.dackar.RCA.ner.hybrid_ner.consolidator

Classes

SpanConsolidatorPolicy	Policy knobs for candidate consolidation.
SpanConsolidator	Consolidates raw candidates into a cleaner set.

Module Contents

class src.dackar.RCA.ner.hybrid_ner.consolidator.SpanConsolidatorPolicy

Policy knobs for candidate consolidation.

v0.1 defaults:

• Dedupe identical spans (same start/end)

• Keep nested spans

• Do not do aggressive partial-overlap pruning yet

dedupe_identical_spans: bool = True

keep_nested_spans: bool = True

prefer_longest_on_overlap: bool = False

class src.dackar.RCA.ner.hybrid_ner.consolidator.SpanConsolidator(policy=None)

Consolidates raw candidates into a cleaner set.

Responsibilities:

• Merge duplicates across generators (same offsets)

• Union provenance sources and label hypotheses

• Optional overlap pruning policies (kept minimal for v0.1)

Parameters:

policy (SpanConsolidatorPolicy | None)

policy

consolidate(doc, candidates)

Parameters:

• doc (src.dackar.RCA.ner.hybrid_ner.models.Document)

• candidates (List[src.dackar.RCA.ner.hybrid_ner.models.CandidateSpan])

Return type:

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