# -*- coding: utf-8 -*-
"""Entry points for PyKEEN.
PyKEEN is a machine learning library for knowledge graph embeddings that supports node clustering,
link prediction, entity disambiguation, question/answering, and other tasks with knowledge graphs.
It provides an interface for registering plugins using Python's entrypoints under the
``pykeen.triples.extension_importer`` and ``pykeen.triples.prefix_importer`` groups. More specific
information about how the PyBEL plugins are loaded into PyKEEN can be found in PyBEL's
`setup.cfg <https://github.com/pybel/pybel/blob/master/setup.cfg>`_ under the ``[options.entry_points]``
header.
The following example shows how you can parse/load the triples from a BEL document with the `*.bel` extension.
.. code-block:: python
from urllib.request import urlretrieve
url = 'https://raw.githubusercontent.com/cthoyt/selventa-knowledge/master/selventa_knowledge/small_corpus.bel'
urlretrieve(url, 'small_corpus.bel')
# Example 1A: Make triples factory
from pykeen.triples import TriplesFactory
tf = TriplesFactory(path='small_corpus.bel')
# Example 1B: Use directly in the pipeline, which automatically invokes training/testing set stratification
from pykeen.pipeline import pipeline
results = pipeline(
dataset='small_corpus.bel',
model='TransE',
)
The same is true for precompiled BEL documents in the node-link format with the `*.bel.nodelink.json` extension and
the pickle format with the `*.bel.pickle` extension.
The following example shows how you can load/parse the triples from a BEL document stored in BEL Commons using the
``bel-commons`` prefix in combination with the network's identifier.
.. code-block:: python
# Example 2A: Make a triples factory
from pykeen.triples import TriplesFactory
# the network's identifier is 528
tf = TriplesFactory(path='bel-commons:528')
# Example 1B: Use directly in the pipeline, which automatically invokes training/testing set stratification
from pykeen.pipeline import pipeline
results = pipeline(
dataset='bel-commons:528',
model='TransR',
)
Currently, this relies on the default BEL Commons service provider at https://bel-commons-dev.scai.fraunhofer.de,
whose location might change in the future.
"""
import numpy as np
from .bel_commons_client import from_bel_commons
from .gpickle import from_pickle
from .nodelink import from_nodelink_file
from .triples import to_triples
__all__ = [
"get_triples_from_bel",
"get_triples_from_bel_nodelink",
"get_triples_from_bel_pickle",
"get_triples_from_bel_commons",
]
[docs]def get_triples_from_bel(path: str) -> np.ndarray:
"""Get triples from a BEL file by wrapping :func:`pybel.io.tsv.api.get_triples`.
:param path: the file path to a BEL Script
:return: A three column array with head, relation, and tail in each row
"""
from pybel import from_bel_script
return _from_bel(path, from_bel_script)
[docs]def get_triples_from_bel_nodelink(path: str) -> np.ndarray:
"""Get triples from a BEL Node-link JSON file by wrapping :func:`pybel.io.tsv.api.get_triples`.
:param path: the file path to a BEL Node-link JSON file
:return: A three column array with head, relation, and tail in each row
"""
return _from_bel(path, from_nodelink_file)
[docs]def get_triples_from_bel_pickle(path: str) -> np.ndarray:
"""Get triples from a BEL pickle file by wrapping :func:`pybel.io.tsv.api.get_triples`.
:param path: the file path to a BEL pickle file
:return: A three column array with head, relation, and tail in each row
"""
return _from_bel(path, from_pickle)
[docs]def get_triples_from_bel_commons(network_id: str) -> np.ndarray:
"""Load a BEL document from BEL Commons by wrapping :func:`pybel.io.tsv.api.get_triples`.
:param network_id: The network identifier for a graph in BEL Commons
:return: A three column array with head, relation, and tail in each row
"""
return _from_bel(str(network_id), from_bel_commons)
def _from_bel(path, bel_importer) -> np.ndarray:
graph = bel_importer(path)
triples = to_triples(graph)
return np.array(triples)