An Approach to Speed-up the Word Sense Disambiguation Procedure through Sense Filtering

TL;DR

This paper introduces a method to accelerate Word Sense Disambiguation by filtering senses through Part-of-Speech Tagging, significantly reducing execution time and improving accuracy.

Contribution

It proposes a novel approach combining POS tagging with sense filtering to speed up WSD, using Bigram approximation and WordNet for efficient disambiguation.

Findings

Execution time reduced by approximately 50%

Improved accuracy in sense disambiguation cases

Effective sense filtering based on POS tags

Abstract

In this paper, we are going to focus on speed up of the Word Sense Disambiguation procedure by filtering the relevant senses of an ambiguous word through Part-of-Speech Tagging. First, this proposed approach performs the Part-of-Speech Tagging operation before the disambiguation procedure using Bigram approximation. As a result, the exact Part-of-Speech of the ambiguous word at a particular text instance is derived. In the next stage, only those dictionary definitions (glosses) are retrieved from an online dictionary, which are associated with that particular Part-of-Speech to disambiguate the exact sense of the ambiguous word. In the training phase, we have used Brown Corpus for Part-of-Speech Tagging and WordNet as an online dictionary. The proposed approach reduces the execution time upto half (approximately) of the normal execution time for a text, containing around 200 sentences. Not only that, we have found several instances, where the correct sense of an ambiguous word is found for using the Part-of-Speech Tagging before the Disambiguation procedure.