Exploring Twitter news biases using urdu-based sentiment lexicon

Abstract

Social media has become a tremendous success in recent times. It has enabled people to keep in touch with each other anytime and anywhere around the globe. People share their opinions and experiences publically through such platforms. Twitter is one of the significant social networking platform that is used by many news media to disseminate breaking news instantaneously. Sentiment analysis of social media is successfully used to gain insights regarding collective behavior of the society. In addition, sentiment analysis is used to detect positive or negative content posted on the social media for various purposes such as riots detection. In this paper, we focus on the sentiment analysis of news tweets in Urdu language by major news sources in Pakistan. By gathering tweets data over the period of 10 months, we built a sentiment lexicon in Urdu language. Moreover, we devise an algorithm that classifies Urdu text into positive, negative, or neutral classes based on the cumulative sentiment score of the text. Our sentiment analysis algorithm achieves 77% accuracy. Furthermore, we have done perspective analysis in which we estimated the bias in the news reporting through tweets with respect to the government with 77.45% accuracy.