Evaluating Knowledge-Based Semantic Measures on Arabic
Abdulgabbar Saif, Mohd Juzaiddin Ab Aziz, Nazlia Omar · International Journal on Communications Antenna and Propagation (IRECAP) · 2014
Semantic measures have received a wide attention from researchers to handle various issues in the different tasks of the computational linguistics and information retrieval. In this paper, we experimentally investigate the performances of the knowledge-based semantic measures on the Arabic language. The state-of-the-art semantic measures are adapted to two knowledge sources: a highly structured source (Arabic WordNet) and semi-structured source (Arabic Wikipedia). The performance of the different semantic measures is evaluated on four Arabic benchmark data sets of the word-to-word semantic similarity/relatedness task. The evaluation results show that Wikipedia is a competitive and promising knowledge source in terms of its high degree of coverage and the variety of the extractable semantic features.