WikidataVisualRepresentation

class WikidataVisualRepresentation(wikidata_ids: Sequence[str], max_id: int | None = None, image_kwargs: Mapping[str, Any] | None = None, **kwargs)

Bases: BackfillRepresentation

Visual representations obtained from Wikidata and encoded with a vision encoder.

If no image could be found for a certain Wikidata ID, a plain (trainable) embedding will be used instead.

Example usage

"""Example of using visual representations from Wikidata."""

from pykeen.datasets import get_dataset
from pykeen.models import ERModel
from pykeen.nn import WikidataVisualRepresentation
from pykeen.pipeline import pipeline

dataset = get_dataset(dataset="codexsmall")
entity_representations = WikidataVisualRepresentation.from_dataset(dataset=dataset)

result = pipeline(
    dataset=dataset,
    model=ERModel,
    model_kwargs=dict(
        interaction="distmult",
        entity_representations=entity_representations,
        relation_representation_kwargs=dict(
            shape=entity_representations.shape,
        ),
    ),
)

Initialize the representation.

Parameters:

• wikidata_ids (Sequence[str]) – The Wikidata IDs.

• max_id (int | None) – The total number of IDs. If provided, must match the length of wikidata_ids.

• image_kwargs (Mapping[str, Any] | None) – Keyword-based parameters passed to pykeen.nn.vision.cache.WikidataImageCache.get_image_paths().

• kwargs – Additional keyword-based parameters passed to pykeen.nn.vision.representation.VisualRepresentation.

Raises:

ValueError – If the max_id does not match the number of Wikidata IDs.

Methods Summary

from_dataset(dataset[, for_entities])	Prepare representations from a dataset.
from_triples_factory(triples_factory[, ...])	Prepare a visual representations for Wikidata entities from a triples factory.

Methods Documentation

classmethod from_dataset(dataset: Dataset, for_entities: bool = True, **kwargs) → Self

Prepare representations from a dataset.

Parameters:

• dataset (Dataset) – The dataset; needs to have Wikidata IDs as entity names.

• for_entities (bool) – Whether to create the initializer for entities (or relations).

• kwargs – Additional keyword-based arguments passed to pykeen.nn.vision.representation.WikidataVisualRepresentation.

Returns:

A visual representation from the training factory in the dataset.

Raises:

TypeError – If the triples factory does not provide labels.

Return type:

Self

classmethod from_triples_factory(triples_factory: TriplesFactory, for_entities: bool = True, **kwargs) → Self

Prepare a visual representations for Wikidata entities from a triples factory.

Parameters:

• triples_factory (TriplesFactory) – The triples factory.

• for_entities (bool) – Whether to create the initializer for entities (or relations).

• kwargs – Additional keyword-based arguments passed to pykeen.nn.vision.representation.WikidataVisualRepresentation.

Returns:

A visual representation from the triples factory.

Return type:

Self