Recomendação de Tags a Partir de Métricas de Qualidade de Atributos Textuais em Aplicações da Web 2.0

Fabiano BelémEder MartinsJussara AlmeidaMarcos GonçalvesGisele L. Pappa

Several popular Web 2.0 applications allow users to assign keywords (tags) to their objects, in order to provide a better description and organization of the shared content. Tag recommendation services may assist users in that task, improving the quality of the available content and, thus, the effectiveness of tag-based Information Retrieval (IR) services, such as searching and classification. This work addresses the task of recommending high quality tags by exploiting not only previously assigned tags, but also terms extracted from other textual features (e.g., title and description) associated with the target object. These sources are exploited to generate candidate terms for recommendation and to compute heuristic metrics that estimate the quality of a candidate. In particular, we propose a heuristic function that combines multiple metrics to produce a final ranking of the recommended tags. We also investigate the use of Genetic Programming (GP) as a tool to generate and evolve ranking functions. We evaluate both the heuristic and GP-based functions in various scenarios, for three popular Web 2.0 applications, namely, LastFM, YahooVideo and YouTube. Our experiments show that our heuristic and GP-based functions significantly outperform state-of-the-art tag recommendation algorithms.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: