Source Impact and Credibility Assessment on Twitter Users

Abstract

Due to the improvement of the internet, several platforms such as Twitter, Facebook, LinkedIn, Instagramwere very popular. They were attracted by the people as the mass media platform’s cost is very high. Because of this popularity, most of the users rely on the information published on social platforms. The problem is ensuring their reliability; what we read is not fake. Credibility is a major issue when dealing with online social media platforms. The focus of this study is measuring user credibility based on the tweets published by each user. In this study, we compare an approach called Credibility Outcome (CREDO) which aims at marking the credibility of an article in an open domain setting, to create a credibility assessment model for Twitter users. CREDO approach consists of various modules to capture the features responsible for the credibility of unstructured texts such as Semantic similarity of articles, Sentiment conveys by the article, Information source credibility, and Keyword extraction value. As tweet is also unstructured text, use CREDO algorithm to measure Twitter user credibility based on the above features and experiment on Twitter dataset reveals that CREDO outperforms the state-of-the-art approaches based on linguistic features.