An insight into semantic similarity aspects using WordNet

Abstract

Semantic similarity is becoming a generic issue in variety of applications in areas of information retrieval, computational linguistic and AI, both in the academia and industry. Examples include: computing semantic similarity, word sense disambiguation, text segmentation, multimodal document retrieval, image retrieval, etc. However, semantic similarity measures have been used showing mixed chances of success. The basic problem is that if semantic measures are used bluntly without understanding, they might decrease retrieval efficiency. There is a need to investigate semantic similarity approaches in order to have better understanding of these approaches. Several semantic methods for determining semantic similarity between terms have been proposed in the literature and most of them have been tested on WordNet. In this paper, we investigate the approaches to compute semantic similarity by mapping word concepts to WordNet ontology and by examining their relationship in that ontology. The paper then provides specific examples for explaining these approaches Further, the paper categorises and compares various approaches for measuring semantic similarity using WordNet ontology.