Uma abordagem para o uso do conceito de Folkauthority em sistemas de recuperação de informa

Abstract

The information overload in the Web 2.0 environment makes the task of retrieving information with quality and relevance quite dicult, since users are not able to examine and interpret a significant part of the information available. Several authors relate the quality and relevance of information available on the Web with the concept of cognitive authority stating that entities possessing expertise in particular subject tend to publish information containing these two features. Therefore, this dissertation considers the hypothesis that the use of a ranking scheme that takes into account the cognitive authority of information sources produces an Information Retrieval System that presents search results containing information of higher quality and relevance to users. Was adopted an approach called Folkauthority to verify this hypothesis { a framework in which the sources of information in the system are categorized with tags according to their cognitive authority { and proposed an architecture for information retrieval whose ranking scheme, called AuthorityRank, is based on the categorization of authorities. The concessions of authority among users were simulated based on data from a real social network and an Information Retrieval System whose ranking scheme considers the framework established by the Folkauthority approach was implemented. This scheme was then compared with other schemes using the metric of NDCG, by which was possible to verify a statistically significant improvement in the quality and relevance of information retrieved through the AuthorityRank scheme when compared with the tf-idf scheme, thus, conforming the validity of the hypothesis with respect to the proposed approach.