9th International Conference on ICT for Sustainable Development

Authors - Kiran V. Sonkamble, Saroj S. Date, Sachin N.Deshmukh
Abstract - Sentiment analysis is a method in the processing of natural language that uses machine learning to detect and extract the polarity of sentiments represented in a text, including positive, negative, or neutral. Due to the increasing acceptance of social media and the increased use of Indian languages, there has been an increase in interest in sentiment analysis in Indian languages in recent years. Machine learning is utilized to analyze and categorize subjective data, such as opinions, sentiments, and attitudes, expressed in text. This approach has been used in a variety of languages, such as Telugu, Marathi, and Bengali. Within this particular context, the evaluation of the models was evaluated in Bengali language including comparing their accuracy, precision, recall, and F1 score. A comparison was made with models applied to different languages, and it was seen that both KNN and SVM consistently performed well across all three languages. However, the presentation of the LR model differed. In phrase of F1 ratings, KNN and LR performed the best in Bengali compared to Telugu, whereas SVM obtained the highest F1 score. Among the Marathi language models, KNN had the greatest F1 score, while SVM achieved the highest recall score. Hence, it is essential to carefully choose the suitable machine-learning algorithm for each language in order to get the utmost precision in sentiment analysis. The sentiment classification step involves categorizing the sentiment expressed in the text using machine learning techniques. For sentiment analysis in Indian regional languages, machine learning techniques such as KNN, Support Vector Machines (SVM), and Logistic regression have been applied. Logistic regressions have demonstrated superior accuracy compared to KNN and SVMs. Accuracy, precision, recall, and F1 score are essential performance analysis approaches to evaluate. These indicators can aid in measuring the effective the algorithms are at identifying sentiment in text.