Twitter Sentiment Analysis using Collaborative Multi Layer Perceptron(MLP) Classifier

Abstract

In the last decade, Twitter is facing a major challenge in identifying the sentiments of the tweets. These tweets are useful in understanding the opinion of different end users about a variety of topics. Sentiment Analysis is more important to analyse the text data in the dataset. But the tweets consist of non-useful characters which make sentiment analysis difficult. With multi-layer perceptron classifier, classification for new tweets as either positive or negative is done. This can be done by using a range of features like tokenisation, lemmatisation, stemming on the words used in the tweets, the sentiment of the words, and the user’s profile. Feature extraction is made to clean the words/phrases containing non-useful characters like URL, Numeric alphabets, punctuations etc. It improves the accuracy for the large amount of data with high efficiency.