minimel.run module

minimel.run.vectorize_text(texts, vectorizer=None, dim=None)

minimel.run.get_scores(golds, preds, per_name=False)

class minimel.run.MiniNED(dawgfile: Path, candidatefile: Path | None = None, modelfile: Path | None = None, vectorizer: Path | None = None, ent_feats_csv: Path | None = None, lang: str | None = None, fallback: Path | None = None)

Bases: object

Parameters:

• dawgfile (Path)

• candidatefile (Optional[Path])

• modelfile (Optional[Path])

• vectorizer (Optional[Path])

• ent_feats_csv (Optional[Path])

• lang (Optional[str])

• fallback (Optional[Path])

predict(text: str, name: str, upperbound=None, all_scores=False)

Make NED prediction.

Parameters:

• text (str) – Some text

• name (str) – An entity name in text

Keyword Arguments:

• all_scores – Output all candidate scores

• upperbound – Create upper bound on performance

Returns:

Wikidata ID

minimel.run.run(dawgfile: Path, candidatefile: Path | None = None, modelfile: Path | None = None, *runfiles: Path, outfile: Path | None = None, vectorizer: Path | None = None, ent_feats_csv: Path | None = None, lang: str | None = None, fallback: Path | None = None, evaluate: bool = False, evalfile: Path | None = None, evalfile_per_name: Path | None = None, predict_only: bool = True, all_scores: bool = False, upperbound: bool = False, split: int | None = None, fold: int | None = None)

Perform entity disambiguation

Parameters:

• dawgfile (Path) – DAWG trie file of Wikipedia > Wikidata count

• candidatefile (Optional[Path]) – Candidate {name -> [ID]} json

• modelfile (Optional[Path]) – Vowpal Wabbit model

• runfiles (Path) – Input file (- or absent for standard input). TSV rows of (ID, {name -> ID}, text) or ({name -> ID}, text) or (text)

• outfile (Optional[Path])

• vectorizer (Optional[Path])

• ent_feats_csv (Optional[Path])

• lang (Optional[str])

• fallback (Optional[Path])

• evaluate (bool)

• evalfile (Optional[Path])

• evalfile_per_name (Optional[Path])

• predict_only (bool)

• all_scores (bool)

• upperbound (bool)

• split (Optional[int])

• fold (Optional[int])

Keyword Arguments:

• outfile – Write outputs to file (default: stdout)

• vectorizer – Scikit-learn vectorizer .pickle or Fasttext .bin word embeddings. If unset, use HashingVectorizer.

• ent_feats_csv – CSV of (ent_id,space separated feat list) entity features

• fallback – Additional fallback deterministic name -> ID json

• evaluate – Report evaluation scores instead of predictions

• evalfile – Write evaluation results to file

• evalfile_per_name – Write evaluation results per name to file

• predict_only – Only print predictions, not original text

• all_scores – Output all candidate scores

• upperbound – Create upper bound on performance

• split – Split the data into several parts

• fold – Use only this fold of the split data

minimel.run.evaluate(goldfile: Path, *predfiles: Path, agg: List[Path] = (), evalfile: Path | None = None)

Evaluate predictions

Parameters:

• gold – Gold data TSV

• pred – Prediction TSVs

• goldfile (Path)

• predfiles (Path)

• agg (List[Path])

• evalfile (Optional[Path])

Keyword Arguments:

• agg – Aggregation jsons (TODO: depend on data…?)

• evalfile – Write evaluation results to file