# -*- coding: utf-8 -*-
"""The **Co**mpletion **D**atasets **Ex**tracted from Wikidata and Wikipedia (CoDEx) datasets from [safavi2020]_.
- GitHub Repository: https://github.com/tsafavi/codex
- Paper: https://arxiv.org/abs/2009.07810
"""
from docdata import parse_docdata
from .base import UnpackedRemoteDataset
BASE_URL = 'https://raw.githubusercontent.com/tsafavi/codex/master/data/triples/'
SMALL_VALID_URL = f'{BASE_URL}/codex-s/valid.txt'
SMALL_TEST_URL = f'{BASE_URL}/codex-s/test.txt'
SMALL_TRAIN_URL = f'{BASE_URL}/codex-s/train.txt'
MEDIUM_VALID_URL = f'{BASE_URL}/codex-m/valid.txt'
MEDIUM_TEST_URL = f'{BASE_URL}/codex-m/test.txt'
MEDIUM_TRAIN_URL = f'{BASE_URL}/codex-m/train.txt'
LARGE_VALID_URL = f'{BASE_URL}/codex-l/valid.txt'
LARGE_TEST_URL = f'{BASE_URL}/codex-l/test.txt'
LARGE_TRAIN_URL = f'{BASE_URL}/codex-l/train.txt'
# If GitHub ever gets upset from too many downloads, we can switch to
# the data posted at https://github.com/pykeen/pykeen/pull/154#issuecomment-730462039
[docs]@parse_docdata
class CoDExSmall(UnpackedRemoteDataset):
"""The CoDEx small dataset.
---
name: CoDEx (small)
citation:
author: Safavi
year: 2020
link: https://arxiv.org/abs/2009.07810
github: tsafavi/codex
statistics:
entities: 2034
relations: 42
training: 32888
testing: 1828
validation: 1827
triples: 36543
"""
def __init__(self, create_inverse_triples: bool = False, **kwargs):
"""Initialize the `CoDEx <https://github.com/tsafavi/codex>`_ small dataset from [safavi2020]_.
:param create_inverse_triples: Should inverse triples be created? Defaults to false.
:param kwargs: keyword arguments passed to :class:`pykeen.datasets.base.UnpackedRemoteDataset`.
"""
# GitHub's raw.githubusercontent.com service rejects requests that are streamable. This is
# normally the default for all of PyKEEN's remote datasets, so just switch the default here.
kwargs.setdefault('stream', False)
super().__init__(
training_url=SMALL_TRAIN_URL,
testing_url=SMALL_TEST_URL,
validation_url=SMALL_VALID_URL,
create_inverse_triples=create_inverse_triples,
**kwargs,
)
[docs]@parse_docdata
class CoDExMedium(UnpackedRemoteDataset):
"""The CoDEx medium dataset.
---
name: CoDEx (medium)
citation:
author: Safavi
year: 2020
link: https://arxiv.org/abs/2009.07810
github: tsafavi/codex
statistics:
entities: 17050
relations: 51
training: 185584
testing: 10311
validation: 10310
triples: 206205
"""
def __init__(self, create_inverse_triples: bool = False, **kwargs):
"""Initialize the `CoDEx <https://github.com/tsafavi/codex>`_ medium dataset from [safavi2020]_.
:param create_inverse_triples: Should inverse triples be created? Defaults to false.
:param kwargs: keyword arguments passed to :class:`pykeen.datasets.base.UnpackedRemoteDataset`.
"""
kwargs.setdefault('stream', False) # See comment in CoDExSmall
super().__init__(
training_url=MEDIUM_TRAIN_URL,
testing_url=MEDIUM_TEST_URL,
validation_url=MEDIUM_VALID_URL,
create_inverse_triples=create_inverse_triples,
**kwargs,
)
[docs]@parse_docdata
class CoDExLarge(UnpackedRemoteDataset):
"""The CoDEx large dataset.
---
name: CoDEx (large)
citation:
author: Safavi
year: 2020
link: https://arxiv.org/abs/2009.07810
github: tsafavi/codex
statistics:
entities: 77951
relations: 69
training: 551193
testing: 30622
validation: 30622
triples: 612437
"""
def __init__(self, create_inverse_triples: bool = False, **kwargs):
"""Initialize the `CoDEx <https://github.com/tsafavi/codex>`_ large dataset from [safavi2020]_.
:param create_inverse_triples: Should inverse triples be created? Defaults to false.
:param kwargs: keyword arguments passed to :class:`pykeen.datasets.base.UnpackedRemoteDataset`.
"""
kwargs.setdefault('stream', False) # See comment in CoDExSmall
super().__init__(
training_url=LARGE_TRAIN_URL,
testing_url=LARGE_TEST_URL,
validation_url=LARGE_VALID_URL,
create_inverse_triples=create_inverse_triples,
**kwargs,
)
def _main():
for cls in [CoDExSmall, CoDExMedium, CoDExLarge]:
d = cls()
d.summarize()
if __name__ == '__main__':
_main()