Sentiment Analysis of Enterprise Mashups Using Scikit and NLTK

Abstract

With the advent of internet and its related technologies, storage and use of data is increasing every second at voluminous speed. There is a need to process this huge volume of data automatically and to find out the required hidden patterns in it. The main objective of our research is to find the covert sentiments in Twitter status messages which can help companies to realize the true potential of their businesses and can also help individuals to make better decision during a purchase. The status messages (tweets) are extracted from twitter about different companies, products, and personalities and are analyzed to find out what are the opinions (positive or negative) of people about those entities? The data is extracted using twitter API. A framework is proposed which facilitates to test different techniques and find one which works best with polarity detection of twitter messages. Bigram, unigram, term frequency and inverse document frequency (TF-IDF) and other feature selection methods are used along with three machine learning algorithms: support vector machines, stochastic decent gradient, and Naïve Bayes. The work also includes performance comparison of different learning methods to analyze which kinds of techniques work best with twitter corpus