A Hybrid Bat Algorithm Based on Combined Semantic Measures for Word Sense Disambiguation

Abstract

The task of assigning an appropriate sense for an ambiguous word depending on its context is referred to as Words Sense Disambiguation (WSD). The objective of WSD is to attain an improved accuracy for real world applications, such as in information extraction, automatic summarisation, or machine translation. The WSD is solved through the implementation of a computational intelligence approach known as the Bat Algorithm (BA). The BA has the potential to explore an expansive area of the search space as it is a population-based algorithm, which makes it considerably efficient in the diversification process. To further improve the search, a local search algorithm referred to as Hill Climbing (HC) is applied that balances the exploration as well as the exploitation aspects. The suggested algorithm has the ability to optimise the semantic value of the words from the inputted text. In this study, the semantic measure depends on the Leacock and Chodorow (LCH) algorithms and the extended Lesk (eLesk). The recommended algorithm is tested based on certain benchmark datasets. According to the experimental results, it is found that our algorithm can derive better quality performance than that of other relevant algorithms. It is, therefore, concluded that the method suggested in this study provides an effectual solution for the WSD problem.