Multilevel User Credibility Assessment in Social Networks (2309.13305v1)

Abstract: Online social networks are one of the largest platforms for disseminating both real and fake news. Many users on these networks, intentionally or unintentionally, spread harmful content, fake news, and rumors in fields such as politics and business. As a result, numerous studies have been conducted in recent years to assess the credibility of users. A shortcoming of most of existing methods is that they assess users by placing them in one of two categories, real or fake. However, in real-world applications it is usually more desirable to consider several levels of user credibility. Another shortcoming is that existing approaches only use a portion of important features, which downgrades their performance. In this paper, due to the lack of an appropriate dataset for multilevel user credibility assessment, first we design a method to collect data suitable to assess credibility at multiple levels. Then, we develop the MultiCred model that places users at one of several levels of credibility, based on a rich and diverse set of features extracted from users' profile, tweets and comments. MultiCred exploits deep LLMs to analyze textual data and deep neural models to process non-textual features. Our extensive experiments reveal that MultiCred considerably outperforms existing approaches, in terms of several accuracy measures.