To search, Click below search items.


All Published Papers Search Service


Experimental Study of Semantic Similarity Measures on Arabic WordNet


Nababteh Mohammed, Deri Mohammed


Vol. 17  No. 2  pp. 131-140


There are several semantic similarity measures that have been used to measure and quantify how much two concepts are alike. However, these measures have been tested, verified and compared in English language, using WordNet (WN). Few concerns have been given to study the impacts of traditional semantic similarity measures on Arabic language, embodied in Arabic WordNet (AWN). This paper aims at investigating the ability of applying semantic similarity measures over AWN and their applicability on Arabic-related applications. Having semantic measures for Arabic language will support many Arabic-based natural language processing applications. In this paper the experimental study was applied on seven semantic similarity measures from numerous semantic similarity measures. The experiments show that Wup measure has achieved the highest correlation with human ratings and the lowest value of MSE. This indicates that the Wup measure has the best performance in calculating the similarity of Arabic word pairs using AWN ontology against the other measures. In the other hand, path measure has the worst performance, because of the lowest correlation with human ratings and the highest value of MSE that it has achieved.


Semantic similarity, semantic similarity measure, WordNet, Arabic WordNet (AWN), AWSS.