Learning Content Trust Based on Two Level Factoid Ranking Model

* This work is partially supported by the National Natural Science Foundation of China under grant of 60673157, the Ministry of Education key project under grant of 105071 and SEC E-Institute: Shanghai High Institutions Grid under grant of 200301. Abstract Trust is an integral component in many kinds of human interaction, allowing people to act under uncertainty and with the risk of negative consequences. While in computer science, most prior work focuses on entity-centered issues such as authentication and reputation, it does not model the content, i.e. the nature and use of the information being exchanged. This paper discusses content trust as a factoid learning problem (factoid here refers to something which can reflect the truth of the content), which extracts factoid from the content and then rank them according to their likehood as trustworthy ones. Learning methods for performing factoid ranking are proposed in this paper, which formalize the problem as ordinal regression or ranking. Trust features for judging the trustworthiness of a factoid is given, and features for constructing the Ranking SVM models are defined. We employ a two level ranking SVM as the implementation of the problem. Finally the evaluating of the model and the experimental results were presented.