src.dackar.RCA.ner.hybrid_ner.pipeline

Attributes

MECH_KEYWORDS

Classes

HybridNERPipeline

Orchestrates the Hybrid NER steps:

Module Contents

src.dackar.RCA.ner.hybrid_ner.pipeline.MECH_KEYWORDS

class src.dackar.RCA.ner.hybrid_ner.pipeline.HybridNERPipeline(schema, generators, consolidator=None, disambiguator=None, classifier=None, compatibility=None, postprocessor=None, desc_gen=None, llm_disambiguator=None, enable_split_component_mechanism=True)

Orchestrates the Hybrid NER steps:

1. Candidate generation (high recall): multiple generators

2. Consolidation: dedupe/merge candidates

3. Rule disambiguation: cheap context guards (optional)

4. ML classifier: span classification (optional)

5. Compatibility engine: enforce group schema rules and multi-label policy

6. Postprocess: flatten entities and provide audit trail

This pipeline is designed so each stage can be swapped or extended without breaking downstream code.

Parameters:

• schema (src.dackar.RCA.ner.hybrid_ner.schema.SchemaIndex)

• generators (List[src.dackar.RCA.ner.hybrid_ner.generators.base.CandidateGenerator])

• consolidator (src.dackar.RCA.ner.hybrid_ner.consolidator.SpanConsolidator | None)

• disambiguator (src.dackar.RCA.ner.hybrid_ner.disambiguation.RuleDisambiguator | None)

• classifier (src.dackar.RCA.ner.hybrid_ner.classifier.SpanClassifier | None)

• compatibility (src.dackar.RCA.ner.hybrid_ner.compatibility.CompatibilityEngine | None)

• postprocessor (src.dackar.RCA.ner.hybrid_ner.postprocess.Postprocessor | None)

• desc_gen (object | None)

• enable_split_component_mechanism (bool)

schema

generators

consolidator

disambiguator

classifier

compatibility

postprocessor

desc_gen = None

llm_disambiguator = None

enable_split_component_mechanism = True

run(doc)

Parameters:

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

Return type:

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