LDS: Java Library for Linked Open Data Based Similarity Measures
Résumé
Semantic similarity measures are metrics that analyze knowledge sources such as ontologies to produce a similarity score between compared concepts. They are used in different information retrieval domains such as semantic search or recommendation systems. Semantic Measure Libraries (SML) on the other hand, are tools that facilitate the use and implementation of semantic similarity measures. Existing libraries are in general dedicated to ontology-based semantic similarity measures. However, with the emergence of Linked Open Data (LOD), which provides a large and diverse source of information on the Web, new LOD-based semantic similarity measures are proposed. These measures need to handle new kinds of data, and existing SML are not designed to support these functionalities. In this paper, we present LDS (Linked Data Similarity) Library, a Java software library of LOD-based semantic similarity measures. LDS implements some well-known similarity measures and provides efficient utilities and tools for similarity calculation and implementation. We conduct a set of experiments to evaluate LDS efficiency. Also, we present a use case where we propose a new similarity measure that extends an existing measure by reusing its components.