GLiNERGrounder

class GLiNERGrounder(matcher: Matcher[R], *, model: str | gliner.GLiNER | None = None, labels: list[str], threshold: float | None = None)[source]

Bases: Grounder[R], WrappedMatcher[R], Generic[R]

An annotator that works via gliner.

Create a grounder based on a pre-defined matcher and a gliner NER model.

Parameters:

  • matcher – A pre-defined matcher

  • model – The name of a gliner model. See here for a list of models. If not given, defaults to GLINER_DEFAULT.

  • labels –

    The labels for entity types.

    Warning

    These aren’t specifically defined anywhere in the GLiNER docs, so you’ll have to experiment with them

  • threshold – The score threshold for predictions. Defaults to 0.5 if not given.

In the following example, a GLiNER grounder is instantiated using an underlying Gilda matcher, which incorporates the disease branch of Medical Subject Headings (MeSH).

import spacy
from ssslm import GildaMatcher, GLiNERGrounder

matcher = GildaMatcher.default()
grounder = GLiNERGrounder(
    matcher=matcher,
    model="urchade/gliner_medium-v2.1",
    labels=["disease", "protein", "gene"],
)
annotations = grounder.annotate(
    "The APOE e4 mutation is correlated with risk for Alzheimer's disease."
)

Methods Summary

annotate(text, **kwargs)

Annotate the text the GLiNER annotator and the wrapped matcher.

Methods Documentation

annotate(text: str, **kwargs: Any) → list[Annotation][source]

Annotate the text the GLiNER annotator and the wrapped matcher.