Neural networks for sentiment analysis on Twitter

Abstract

The online medium has become a significant way that people express their opinions online. Sentiment analysis can be used to find out the polarity of an opinion, such as positive, negative, or neutral. Sentiment analysis has applications such as companies getting their customer's opinions on their products, political sentiment analysis, or opinions on movie reviews. Recent research has involved looking at text from online blogs, tweets, online movie reviews, etc. to try and classify the text as being positive, negative, or neutral. For this research, a feedforward neural network will be experimented with for sentiment analysis of tweets. The training set of tweets are collected using the Twitter API using positive and negative keywords. The testing set of tweets are collected using the same positive and negative keywords.