Text sentiment analysis of Marathi language in English And Devanagari script

No Thumbnail Available
Gavali, Harry
Issue Date
MSc in Data Analytics
Dublin Business School
Items in eSource are protected by copyright. Previously published items are made available in accordance with the copyright policy of the publisher/copyright holder.
Marathi is a language spoken by a very large number of people in India and about 10% of the Indian population uses ‘Marathi + English’ while texting one another. This study focuses on text sentiment analysis of the mixed language text of Marathi (written in English) first, and then compares the accuracy again after the same sentences (written in Devanagari script) have been translated using Google’s cloud translation services. Same machine learning techniques were applied on both in order to maintain equality. A new and accurate dataset which comprised of day to day sentences was compiled manually in order to reduce error. The outputs were later compared and the need to develop such researches further is highlighted. The results of the research show that the algorithms like Random Forest and SVM give us the highest accuracies of 65.41% and 64.16% respectively.