Graph Embeddings for Linked Data Clustering
Siham Eddamiri, El Moukhtar Zemmouri, Asmaa Benghabrit · 2018
The availability and accessibility of large RDF data in the Linked Open Data cloud encourage the machine learning community to develop approaches and techniques to extract useful knowledge from such type of data. Moreover, Data Clustering is identified as a crucial task for many web-based applications. In this paper, we present an approach that uses neural language models for RDF data clustering. We, first, generate sequences of entities extracted from several graph substructures using Doc2vec and Word2vec combined with TF-IDF. Then we apply K-Means to cluster generated vectors. Our experiments on real datasets show good results when applying TF-IDF with Doc2vec for vector representation of RDF data.