Recomendação associativa de tags considerando múltiplos atributos textuais

Detalhes bibliográficos
Ano de defesa: 2011
Autor(a) principal: Fabiano Muniz Belem
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de Minas Gerais
UFMG
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://hdl.handle.net/1843/SLSS-8KEEQ3
Resumo: Several Web 2.0 applications allow users to assign keywords (tags) to the content, in order to provide a better organization and description of the shared content. Tag recommendation tasks may assist users in this task, improving the quality of the available information and, thus, the effectiveness of Information Retrieval (IR) services which exploit tags, such as searching and classification. This work addresses two tag recommendation tasks. The first one aims at suggesting relevant tags to a target object. The second task is personalized, aiming at recommending tags which are simultaneously relevant to a target object and a target user.Our tag recommendation strategies focused on the object jointly exploit three dimensions of the problem: (i) term co-occurrence with tags pre-assigned to the target object, (ii) terms extracted from multiple textual features, and (iii) several metrics of tag relevance.Our personalized tag recommendation strategies exploit all dimensions mentioned above and a metric related to the history of tag assignments of the target user. In particular, we propose several new heuristic methods, which extend previous, highly effective and efficient, state-of-the-art strategies by including new metrics that try to capture how accurately a candidate term describes the objects content. We also exploit two learning to rank techniques, namely RankSVM and Genetic Programming, for the task of generating ranking functions that combine multiple metrics to accurately estimate the relevance of a tag toa given object or pair user-object. We evaluate all proposed methods in various scenarios for three popular Web 2.0 applications, namely, LastFM, YouTube and YahooVideo. We found that our heuristics greatly outperform the best baseline in each scenario and task. Some further, modest improvements can also be achieved, in some scenarios of both personalized and object-focused, with our new learning-to-rank based strategies. However, they have the additional advantage of being quite flexible and easily extensible to exploit other aspects of the tag recommendation problem.