{"title":"Sentiment analysis of Mizo using lexical features in low resource based models","authors":"Mercy Lalthangmawii , Thoudam Doren Singh","doi":"10.1016/j.nlp.2025.100181","DOIUrl":null,"url":null,"abstract":"<div><div>Sentiment analysis is a vital area of natural language processing (NLP) for interpreting emotions in user-generated content. Although significant progress has been made for widely spoken languages, low-resource languages such as Mizo remain underexplored. This study addresses this gap by developing the first comprehensive sentiment analysis framework for Mizo language. We created a meticulously annotated data set that captures positive, negative, and neutral sentiments. Using classical machine learning models enhanced with lexicon features and transfer learning with XLM-RoBERTa, we demonstrate the feasibility of sentiment analysis in low-resource settings. Our approach achieves an accuracy of 82% with Logistic Regression and 78% with XLM-RoBERTa, which establishes a benchmark for future research in Mizo sentiment analysis.</div></div>","PeriodicalId":100944,"journal":{"name":"Natural Language Processing Journal","volume":"13 ","pages":"Article 100181"},"PeriodicalIF":0.0000,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Natural Language Processing Journal","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2949719125000573","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Sentiment analysis is a vital area of natural language processing (NLP) for interpreting emotions in user-generated content. Although significant progress has been made for widely spoken languages, low-resource languages such as Mizo remain underexplored. This study addresses this gap by developing the first comprehensive sentiment analysis framework for Mizo language. We created a meticulously annotated data set that captures positive, negative, and neutral sentiments. Using classical machine learning models enhanced with lexicon features and transfer learning with XLM-RoBERTa, we demonstrate the feasibility of sentiment analysis in low-resource settings. Our approach achieves an accuracy of 82% with Logistic Regression and 78% with XLM-RoBERTa, which establishes a benchmark for future research in Mizo sentiment analysis.