Sentiment Analysis from Bengali Depression Dataset using Machine Learning
Md. Rafidul Hasan Khan, Umme Sunzida Afroz, Abu Kaisar Mohammad Masum, Sheikh Abujar, Syed Akhter Hossain
Abstract
Nowadays, Sentiment Analysis is one of the advanced matters of natural language processing. Sentiment analysis determines a particular pole of a paragraph. Our purpose is to find the sentiment from the Bengali paragraph which is happy or sad using various types of machine learning classification analysis algorithms. For doing this we are collecting data from various social network sites, Bengali blogs, etc. To get a compatible result, we passed through many difficulties. Bengali text preprocessing is one of the complex parts of all. After preprocessing the data, we tokenized the data by using Countvectorizer. After that, we applied six different algorithms to predict almost high accuracy. Among them, the Multinomial Naive Bayes provide us the maximum accuracy which is 86.67%.