Analyzing and Improving Accuracy of Tagging using Hybrid Approach ror Gujarati Language

Abstract

Part of speech tagging is allocating a tag or diverse lexical magnificence marker to every phrase in a sentence which is helpful in numerous applications of like phrase disambiguation, information processing, parsing, and language transformation. Thus, tagging is considered the fundamental step for any application of language processing. newlinePeople widely speak Gujarati in Gujarat. The language tool is the strict intention of building an accurate and efficient POS Tagger to recognize the ambiguities of language lexical items. All anticipated taggers have been based on terrific Tag sets, developed using extraordinary organization and person. All proposed POS taggers for Indian languages had been based on an exceptional Tag set, evolved by exclusive organizations and individuals. Our objective in this work is to develop an effective POS tagger for Gujarati. newlineWe have used a BIS tag set that involves eleven main categories of Gujarati tags. We have used Data sets of Entertainment, Arts and Culture, Science, and Sports from Technology Development for Indian Languages (TDIL). We have adopted a statistical approach like Hidden Markov Model (HMM) as the first step for development and then evaluated the rule-based method via applying rules defined by us based on Gujarati linguistics to investigate the accuracy of a part of speech tagger for the Gujarati language. After applying the proposed hybrid approach, a leisure information set has more than 90,000 phrases and improves the accuracy of part of the speech tagger. The hybrid approach works better than HMM and Rule by 10-20% higher accuracy. We also explored comparative analysis of hybrid approach with various deep learning approaches, while experimenting with deep learning improves accuracy by 5% higher than the Hybrid approach. newline