Abstract (eng)
In recent years, since the abundance of published information in social media platforms, research regarding the examination of credibility is becoming increasingly important. Users gather information on such platforms and take part in distribution of content. The body of an information has many facets such as facts, statements, opinions, and sentiment in combination with various types of media. Generally, platforms lack verification of credentials and allow anonymous content to be posted.
While anonymity is important in the matter of freedom of speech these information artifacts can be misused to deliberately spread false content. Consequently, for the the average user it is becoming more difficult to distinguish between false and accurate information.
Therefore, the aim of this thesis is to develop a prototype, which helps to increase the awareness of inaccurate information.
In a first step, we conducted a user study to reveal credibility perception of Twitter posts. Based on the study we identified features to develop automatic credibility classification approaches. To demonstrate the approaches we propose a prototype that allows searching for Twitter posts and filter regarding credibility, sentiment and media type. The proposed prototype enhances the information search process and adds visual cues for calculated sentiment and credibility scores.
The evaluation of the prototype in terms of usability suggests a positive user acceptance within the participating user group.