Estratégias para busca do texto completo de artigos catalogados em uma biblioteca digital

• Autor(a) principal: Allan Jones Costa e Silva
• Data de Publicação: 2007
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/RVMR-794QAJ
• topic: busca em texto; biblioteca digital; Recuperação da informação; Bibliotecas digitais; World Wide Web (Sistema de recuperação da informação); Computação; Sistemas de recuperação da informação

This dissertation proposes a process that uses results from queries submitted to search engines for finding the URL of the corresponding full-text, or of any relevant related material, for those articles cataloged in a digital library for which this information is missing. We present a comprehensive study of this process in different situations by investigating different query strategies applied to three general purpose search engines (Google, Yahoo!, MSN) and two specialized ones (Scholar and CiteSeer), considering five user scenarios characterized by distinct requirement levels. Specifically, we have conducted a set of experiments focused on articles taken from BDBComp - Brazilian Digital Library of Computing and DBLP - Digital Bibliography & Library Project. According to the results of these experiments, Scholar has shown to be more effective than the other tested search engines for this task in all considered scenarios. Moreover, our experiments show that a simple combination Scholar-Google with a re-ranking strategy provides even better results. Our study also presents an analysis of the impact of different factors on the likelihood of finding the full-text of the searched articles.