class WK3l15k(graph_pair='en_de', side='en', cache_root=None, eager=False, create_inverse_triples=False, random_state=0, split_ratios=(0.8, 0.1, 0.1), force=False)[source]

Bases: pykeen.datasets.wk3l.MTransEDataset

The WK3l-15k dataset family.


This dataset contains artifacts from incorrectly treating literals as entities.

Initialize the dataset.

  • graph_pair (str) – The graph-pair within the dataset family (cf. GRAPH_PAIRS).

  • side (str) – The side of the graph-pair, a substring of the graph-pair selection.

  • cache_root (Optional[str]) – The cache root.

  • eager (bool) – Whether to directly load the dataset, or defer it to the first access of a relevant attribute.

  • create_inverse_triples (bool) – Whether to create inverse triples.

  • random_state (Union[None, int, Generator]) – The random state used for splitting.

  • split_ratios (Tuple[float, float, float]) – The split ratios used for splitting the dataset into train / validation / test.

  • force (bool) – Whether to enforce re-download of existing files.


ValueError – If the graph pair or side is invalid.

Attributes Summary


The internal dataset name


The mapping from (graph-pair, side) to triple file name

Attributes Documentation

DATASET_NAME: ClassVar[str] = 'WK3l-15k'

The internal dataset name

FILE_NAMES: ClassVar[Mapping[Tuple[str, str], str]] = {('en_de', 'de'): 'P_de_v6.csv', ('en_de', 'en'): 'P_en_v6.csv', ('en_fr', 'en'): 'P_en_v5.csv', ('en_fr', 'fr'): 'P_fr_v5.csv'}

The mapping from (graph-pair, side) to triple file name