Ir para o conteúdo

Overview

package wittgenstein_rebel_extractor

Wittgenstein REBEL Relation Extractor Library.

Provides parsers and pipeline wrappers to extract structured Knowledge Graph triplets using REBEL.

Classes

  • KGTriplet — Pydantic model representing a Knowledge Graph triplet extracted from text.

  • RebelExtractor — Wraps HuggingFace seq2seq pipeline for local relation extraction using REBEL.

Functions

  • parse_rebel_triplets — Parse raw decoded sequence output from the REBEL model into structured KGTriplet models.

wittgenstein_rebel_extractor.parse_rebel_triplets

mkapi_definition_mkapi parse_rebel_triplets(decoded_text: str) → List[KGTriplet]

Parse raw decoded sequence output from the REBEL model into structured KGTriplet models.

The expected format is

subject1 object1 relation1 object2 relation2 subject2...

wittgenstein_rebel_extractor.KGTriplet

mkapi_definition_mkapi class KGTriplet()

Bases : BaseModel

Pydantic model representing a Knowledge Graph triplet extracted from text.

wittgenstein_rebel_extractor.RebelExtractor

mkapi_definition_mkapi class RebelExtractor(model_name: str = 'Babelscape/rebel-large', device: str = 'cpu')

Wraps HuggingFace seq2seq pipeline for local relation extraction using REBEL.

Methods

wittgenstein_rebel_extractor.RebelExtractor.extract_relations

mkapi_definition_mkapi method RebelExtractor.extract_relations(text: str, max_length: int = 256, length_penalty: float = 0.1, num_beams: int = 3) → List[KGTriplet]

Extract structured relation triplets from text using REBEL.