About the Project
The goal of the three years long project is the creation of new mechanisms for supporting users in the evaluation of Web content credibility and improvement of the quality of Web content credibility evaluation.
Our society increasingly relies on the vast amount of information available on the Internet (and the WWW in particular). As a consequence, the issue of judging or evaluating the credibility of this information becomes of crucial importance. Thus, algorithms and mechanisms for the assessment of credibility of Web content can have a great, positive social impact.
Innovative aspects of RecONCILE include:
Determination of an operational definition of credibility based on research in the social sciences. This definition should be the basis of the implementation and verification of mechanisms for supporting credibility evaluation.
Definition of a methodology for the evaluation of Web content credibility and of the outputs of algorithms for credibility evaluation. This methodology should be based on social experiments and consider the evaluator’s social and individual context. Based on this methodology, the Web Credibility Corpus will be created.
Design of method for finding the relevant and context-aware content to evaluate credibility in (near) real-time. This task involves near real-time crawling for recent web content, finding social context information, fast trust propagation, employing data mining techniques (i.e. collaborative filtering) and fast heuristic rules for discovering relevant evaluation information.
Study of the relation of the amount of information and its freshness with the accuracy of Web content credibility evaluation provided by our mechanism. The analysis of such trade-offs will evaluate the applicability of the approach for various practical applications.
Construction of robust credibility information against various adversarial attacks and other credibility-biasing scenarios. This task involves definition of adversarial models and what-if scenarios and the performance evaluation of the resulting web credibility with various trust models via social simulation and social experiments.
Analysis of the impact of using Web credibility information on online activities: how does identification of trustworthy sites influence the spreading and consumption of Web information, resulting in more effective structure of the online (social) ties among users? Propagating relevant and useful content should be made more efficient through credibility management.
Evaluation of the proposed mechanisms on use-cases that require supporting the credibility evaluation of users in realistic settings, and using the Web Credibility Corpus.
The most important practical results of the project will be a Web Credibility Corpus (a sample collection of Web pages combined with multiple human credibility evaluations that will serve as a benchmark for evaluation or proposed algorithms), a beta version of the software library of credibility evaluation algorithms, and reports from use cases of the proposed mechanism for supporting credibility evaluation. The use cases will test whether the credibility evaluation algorithms included in the library can be used in practice. The use-cases will be focused in three areas: general search, medial information on the Web and open shared knowledge base for physics (in cooperation with CERN).