SENTIMENT ANALYSIS OF NEWS HEADLINE USING MARATHI –SENTIWORD DICTIONARY 2.0

Abstract

Sentiment analysis in regional languages like Marathi is essential for understanding public opinion in local news and media but presents unique challenges due to linguistic limited annotated data. In this research paper work to the two different lexicon Marathi dictionaries, like MSWN 2.0 and MSWN 3.0 varying in the distribution of positive and negative opinions and here find the two different polarities for the MSWN. This is uses datasets sourced from Kaggle, categorized by sentiment, to assess and compare several machine learning classifiers. Include Logistic Regression (LR), Stochastic Gradient Descent (SGD), Support Vector Machine (SVM), Nearest Neighbors (NN), Decision Tree (DT), Naive Bayes (NB), and an Ensemble Classifier .Recognizing accuracy and F-score measures, these are performance tests for each classifier utilizing ever larger dataset sizes. This is done to ascertain which classifiers provide the best consistent performance for Marathi sentiment analysis and reveal scalability over stages wise dataset sizes. With Ensemble Methods obtaining a peak accuracy of 99.71% and F-score of 99.84% in MSWN 2.0, and 97.48% accuracy with an F-score of 97.55% in MSWN 3.0, Ensemble Methods and Decision Trees attained the greatest performance in both MSWN 2.0 and MSWN 3.0. Particularly for models like Logistic Regression, SVM, and Naive Bayes, MSWN 3.0 shown gains in stability. Although SVM and SGD performed comparably in both datasets, MSWN 3.0 produced more consistent results generally; Ensemble Methods is still the most dependable method available.