Skip to content

Latest commit

 

History

History
94 lines (66 loc) · 3.34 KB

README.md

File metadata and controls

94 lines (66 loc) · 3.34 KB
Raw

PyPI Downloads

Section Description
RDF2Vec What is RDF2Vec?
Installing Installing pyRDF2Vec
Getting started A quick introduction
Documentation A link to our documentation
Citation Citing pyRDF2Vec in scholarly articles

What is RDF2Vec?

This repository contains an implementation of the algorithm in "RDF2Vec: RDF Graph Embeddings and Their Applications" by Petar Ristoski, Jessica Rosati, Tommaso Di Noia, Renato De Leone, Heiko Paulheim ([paper] [original code]).

RDF2Vec is an unsupervised technique that builds further on Word2Vec, where an embedding is learned per word by either predicting the word based on its context (Continuous Bag-of-Words (CBOW)) or predicting the context based on a word (Skip-Gram (SG)). To do this, RDF2Vec first creates "sentences" which can be fed to Word2Vec by extracting walks of a certain depth from the Knowledge Graph.

Installation

Few options:

  • (python -m) pip install pyRDF2Vec
  • Clone the repository & run python setup.py install
  • (python -m) pip install git+git://github.com/IBCNServices/pyRDF2Vec.git

Getting Started

Create a Knowledge Graph object

First, you will need to create a Knowledge Graph object (defined in graph.py). We offer several conversion options (such as converting from rdflib or from an endpoint), which can be found in converters.py.

from rdf2vec.converters import rdflib_to_kg

# We want to filter out all triples with certain predicates
label_predicates = [
    'http://dl-learner.org/carcinogenesis#isMutagenic'
]

kg = rdflib_to_kg('sample/mutag.owl', label_predicates=label_predicates)

Define a walking strategy

pyRDF2Vec offers several walking strategies, which can be found in the walkers/ module.

from rdf2vec.walkers import RandomWalker

# We specify the depth and maximum number of walks per entity
random_walker = RandomWalker(4, float('inf'))

Create embeddings

Then, we can create embeddings for a list of entities:

from rdf2vec import RDF2VecTransformer

transformer = RDF2VecTransformer(walkers=[random_walker], sg=1)
# Entities should be a list of URIs that can be found in the KG
embeddings = transformer.fit_transform(kg, entities)

For a more elaborate example, check example.py. You can run it as follows: PYTHONHASHSEED=42 python3 rdf2vec/example.py. The PYTHONHASHSEED is to ensure determinism.

Documentation

Referencing

If you use pyRDF2Vec in a scholarly article, we would appreciate a citation:

@misc{pyrdf2vec,
      title={pyRDF2Vec: A python library for RDF2Vec},
      author={Gilles Vandewiele and Bram Steenwinckel and Michael Weyns
      		  and Pieter Bonte and Femke Ongenae and Filip De Turck},
      year={2020},
      note={\url{https://github.com/IBCNServices/pyRDF2Vec}}
}