Sentiment Analysis of Social Media Contents using Machine Learning Algorithms

Abstract

In natural language processing, sentiment analysis of social media contents is proven to be very effective to analyse huge and complex amount of unstructured data for better decision making. Social media provides an online environment for the users to show their behaviours and emotions through tweets and post etc. Sentiment analysis of any written text especially social media content is applicable to extract the opinions, emotions and meaningful insights. There are many challenges in the accurate sentiment analysis of available social media contents. The challenges can be both technical and theoretical. Several techniques have been suggested in the past but those failed to overcome the mentioned issues in an optimal way because within their work limited datasets were evaluated. The proposed methodology is helpful to overcome above mentioned issues in data acquisition, feature encoding, data pre-processing, feature selection, and classification. In feature encoding phase, a hybrid approach of bi-gram and tri-gram is used for embedding of words. In the experiments, several benchmark datasets have been utilized to measure the effectiveness of the proposed framework. The proposed methodology gives better or at least comparable results with maximum confidence and with less computational complexity. The average accuracy results were in the range from 89-91 with the multilayer perceptron neural network. The mechanism of this work will be helpful to enhance the sentiment analysis process of multifaceted types of social media and blogs contents.