Lexical Ontology-Based Computational Model to Find Semantic Similarity

Abstract

Finding semantic similarity between two words or concepts has been considered as a challenging task in the field of natural language processing. Some lexical ontology-based approaches have been developed for this purpose. However, these approaches have been tested only for English language. Based on survey, there is no computational model for calculating semantic similarity between Hindi concepts. We cannot ignore Hindi language, because it is the third most spoken language of the world. In this paper, we present a computational model for calculating semantic similarity between words/concepts with the help of lexical ontology, which has been tested for Hindi language. Further, experiments have been carried out on a benchmark data set translated from English to Hindi. In our proposed computational model, Hindi WordNet has been used to get relational information between Hindi concepts. Existing popular semantic similarity approaches have been used to calculate semantic similarity. Miller and Charles’s benchmark data set was used to evaluate our proposed approach. We calculated the semantic similarity between 20 word pairs by using three different semantic similarity measures. Accuracy of the results was measured by calculating correlation coefficient between these similarity measures and human judgment. Our proposed model is useful in following ways. Firstly, it allows us to study and analyze the results of available semantic similarity methods on Hindi words. Secondly, it provides a general module along with algorithms, which can be tuned to develop similar modules for any other language.