NEURAL NETWORK BASED APPROACH FOR SENTIMENT CLASSIFICATION
Abstract
Opinions are central to all human activities and are key influencers of our behaviors. The study of sentiment analysis and opinion mining deals with attitude and emotions. Opinion mining has several challenges. The first challenge is that a word is either positive in one situation or negative in another situation. A second challenge is that people can’t express opinions the same way. The online social media make it to use input for their decision making. Therefore, sentiment analysis can be performed using twitter messages. The proposed work is to handle the issues of public health concern. In traditional surveillance systems, observing the public health concern is not only expensive but also suffers from limited coverage and significant delay. To overcome this problem, twitter messages are used which are free of cost and produced in world wide. In this the public concern can be measured using two-step word alignment approach. In the first step, raw reviews are separated into personal reviews and news reviews. In the second step, personal reviews are further classified into Personal Negative and Personal Non-Negative. In both steps, the training data is generated automatically using an emotion-oriented, clue-based method and the trained dataset can be tested using machine learning model such as Naive Bayes. The proposed algorithm will increase the accuracy for epidemic domain.
Author
T.Vijayalakshmi
B.Swaathi
U.Sudarmani
Download
