Credibility Assessment in Microblogs

Join zoom meeting
https://nus-sg.zoom.us/j/81685992057?pwd=am10QUhWbDdQZUpWSUVTbVBBazc2dz09

Abstract:

Microblogging platforms such as Twitter allow users to post publicly viewable short messages. Posts shared by users pertaining to real-world events or themes can provide a rich ``on-the-ground'' live update of the events for the benefit of everyone. Unfortunately, not all posted information may not be credible and the same platform can be exploited to spread misinformation across unsuspecting users on the platform. As such, Twitter has become a prime target for the spread of fake news and misinformation. Despite advances in tools and techniques for the detection and verification of fake news, assessing the credibility of information remains a challenge. In this thesis, we propose to assess the credibility of microblogs at a finer granularity, using evidence from external sources, and considering the relations between the claims.

Existing credibility assessment work have focused on identifying features for discriminating the credibility of messages propagated in Twitter. However, they do not handle tweets that contain multiple pieces of information, each of which may have different level of credibility. Here, we introduce the notion of a ``claim'', and design algorithms to identify claims in a corpus of tweets related to some major event. Extensive experiments on real world datasets show the effectiveness of the proposed approach in identifying claims in various events.

Next, we present an interactive framework called iFACT for assessing the credibility of claims from tweets. The proposed framework collects independent evidence from web search results (WSR). It utilizes features from the search results to determine the probabilities that a claim is credible, not credible or inconclusive. iFACT allows users to be engaged in the credibility assessment process by providing feedback as to whether the web search results are relevant, support or contradict a claim. In addition, iFACT also identifies dependencies between related claims, and use these dependencies to adjust the likelihood estimates of a claim being credible, not credible or inconclusive. Experiment results on multiple real world datasets demonstrate the effectiveness of WSR features and its ability to generalize to claims of new events. Case studies also show the usefulness of claim dependencies and how the proposed approach can give explanation to the credibility assessment process.

Finally, we observe that the credibility of claims depends not just on the content, but also on the time period that the claim is purported to be valid for. We extend our notion of claims to include temporal information, and develop an end-to-end framework for evaluating time-sensitive claims. The framework generates alternate claims and takes into consideration relationships between the alternate claims and the target claim to performs a joint credibility assessment. Experiment results shows the effectiveness of the proposed framework to increase the accuracy of claim assessment.