# largely adapted from pykeen.datasets.ea.openea
import pathlib
from typing import Dict, Literal, Optional, Tuple
from .base import BACKEND_LITERAL, BASE_DATASET_MODULE, ZipEADatasetWithPreSplitFolds
OPEN_EA_MODULE = BASE_DATASET_MODULE.module("open_ea")
# graph pairs
GraphPair = Literal["D_W", "D_Y", "EN_DE", "EN_FR"]
D_W: GraphPair = "D_W"
D_Y: GraphPair = "D_Y"
EN_DE: GraphPair = "EN_DE"
EN_FR: GraphPair = "EN_FR"
GRAPH_PAIRS: Tuple[GraphPair, ...] = (D_W, D_Y, EN_DE, EN_FR)
# graph sizes
GraphSize = Literal["15K", "100K"]
SIZE_15K: GraphSize = "15K"
SIZE_100K: GraphSize = "100K"
GRAPH_SIZES = (SIZE_15K, SIZE_100K)
# graph versions
GraphVersion = Literal["V1", "V2"]
V1: GraphVersion = "V1"
V2: GraphVersion = "V2"
GRAPH_VERSIONS = (V1, V2)
[docs]class OpenEA(ZipEADatasetWithPreSplitFolds):
"""Class containing the OpenEA dataset family.
Published in `Sun, Z. et. al. (2020) A Benchmarking Study of Embedding-based Entity Alignment for Knowledge Graphs <http://www.vldb.org/pvldb/vol13/p2326-sun.pdf>`_,
*Proceedings of the VLDB Endowment*"""
#: The link to the zip file
_FIGSHARE_LINK: str = "https://figshare.com/ndownloader/files/34234391"
#: The hex digest for the zip file
_SHA512: str = (
"c1589f185f86e05c497de147b4d6c243c66775cb4b50c6b41ecc71b36cfafb4c"
"9f86fbee94e1e78a7ee056dd69df1ce3fc210ae07dc64955ad2bfda7450545ef"
)
_GRAPH_PAIR_TO_DS_NAMES: Dict[GraphPair, Tuple[str, str]] = {
"D_W": ("DBpedia", "Wikidata"),
"D_Y": ("DBpedia", "YAGO"),
"EN_DE": ("DBpedia_EN", "DBpedia_DE"),
"EN_FR": ("DBpedia_EN", "DBpedia_FR"),
}
def __init__(
self,
graph_pair: GraphPair = "D_W",
size: GraphSize = "15K",
version: GraphVersion = "V1",
backend: BACKEND_LITERAL = "pandas",
npartitions: int = 1,
use_cache: bool = True,
cache_path: Optional[pathlib.Path] = None,
):
"""Initializes an OpenEA dataset.
:param graph_pair: which pair to use of "D_W", "D_Y", "EN_DE" or "EN_FR"
:param size: what size ("15K" or "100K")
:param version: which version to use ("V1" or "V2")
:param backend: Whether to use "pandas" or "dask"
:param npartitions: how many partitions to use for each frame, when using dask
:param use_cache: whether to use cache or not
:param cache_path: Path where cache will be stored/loaded
:raises ValueError: if unknown graph_pair,size or version values are provided
"""
# Input validation.
if graph_pair not in GRAPH_PAIRS:
raise ValueError(f"Invalid graph pair: Allowed are: {GRAPH_PAIRS}")
if size not in GRAPH_SIZES:
raise ValueError(f"size must be one of {GRAPH_SIZES}")
if version not in GRAPH_VERSIONS:
raise ValueError(f"version must be one of {GRAPH_VERSIONS}")
self.graph_pair = graph_pair
self.size = size
self.version = version
# ensure zip file is present
zip_path = OPEN_EA_MODULE.ensure(
url=OpenEA._FIGSHARE_LINK,
name="OpenEA_dataset_v2.0.zip",
download_kwargs=dict(hexdigests=dict(sha512=OpenEA._SHA512)),
)
inner_cache_path = f"{graph_pair}_{size}_{version}"
inner_path = pathlib.PurePosixPath("OpenEA_dataset_v2.0", inner_cache_path)
actual_cache_path = self.create_cache_path(
OPEN_EA_MODULE, inner_cache_path, cache_path
)
super().__init__(
cache_path=actual_cache_path,
use_cache=use_cache,
zip_path=zip_path,
inner_path=inner_path,
backend=backend,
npartitions=npartitions,
dataset_names=OpenEA._GRAPH_PAIR_TO_DS_NAMES[graph_pair],
)
@property
def _canonical_name(self) -> str:
return f"{self.__class__.__name__}_{self.graph_pair}_{self.size}_{self.version}"
@property
def _param_repr(self) -> str:
return (
f"graph_pair={self.graph_pair}, size={self.size}, version={self.version}, "
)
