Political Opinion Mining Framework (POMF): Framework for Sentiment Analysis of Tweets to Discover Political Sentiments for Popularity Prediction

Abstract

Today, social media is providing a virtual platform for sharing people’s opinion. In fact, it is indispensable for enterprises to consider social feedback for their organic growth prospects. Moreover, the shared sentiments over social media have wherewithal to influence other people in making decisions. This capability of social media has attracted business of different domains. Politics is one such domain where social feedback could change the fate  of political parties. Twitter is the one of the popular micro-blogging website which reflects political views of people with its 330+ millions of users and more than 500 million tweets  per day. The User Generated Content (UGC) over Twitter is rich in sentiments. In this paper, we proposed a deep learning framework known as Political Opinion Mining Framework (POMF) for discovering political sentiments to predict popularity of parties. The framework involves a novel feature selection method that is comprehensive besides optimizing the training process. It exploits Natural Language Processing (NLP) for extra-linguistic contextual information to leverage the quality of training. An improved form of Recurrent Neural Network based on Long Short Term Memory (LSTM) is used in the underlying algorithm named Deep Political Sentiment Discovery (DPSD). The performance of the proposed framework is compared with a many state of the art machine learning algorithms. The empirical study revealed that the DPSD outperforms traditional machine learning techniques.