Filtering Information with imprecise social criteria: A FOAF-based backlink model

Several current approaches to information filtering in Web search engines implement models that use backlinks as a metric of subjective value that complements information retrieval techniques based on the content of the documents. Nonetheless, these models are somewhat “blind” to the reputation or trustworthiness of the creators of the links. The growing increase in interest in applications of social network analysis to the Web provides an opportunity to introduce such social metrics as a complement to existing backlink–based algorithms. In this paper, we describe an approach that uses a FOAF-like vocabulary to derive imprecise degrees of strength on explicitly declared social relations. A backlink model called PeopleRank — that uses the same scheme of the popular PageRank algorithm of the Google search engine — is used as a metric of social relevance. Then, this metric is used to weight the results of the PageRank, obtaining an straightforward “socially weighted” version of the algorithm. Illustrative examples of an implementation using the JUNG Java libraries are also provided.