Corroborating Information from Web Sources

Information available on the Internet is abundant but often inaccurate. Web sources have different degrees of trustworthiness based on their accuracy, freshness, origin or bias. Web users are then left with the daunting task of assessing the correctness of possibly conflicting answers to their queries. In this paper, we present techniques for corroborating information from different web sources. We discuss techniques that estimates the truthfulness of answers and the trustworthiness of the sources based on an underlying probabilistic model. We show how to apply data corroboration to a web setting where data sources can have multiple forms, all with various quality issues: individual web sites, search engine query results, user reviews, map and street view data, and social tags.