Um arcabouço para busca e recomendação de artigos científicos

Abstract

Bibliographical search is an important and a difficult task in the scientific research and it is commonly accomplished by many people around the world. There is currently an increasing number of research papers available on the Web, thus a significant portion of this task is currently performed in the digital environment. Due to this large amount of information, it has become almost impossible to cover all the relevant related work present in the literature. In order to aid researchers to find related papers, recommender systems have been applied to filter and suggest content of interest. The main problem with the current approaches is the use of privileged information, such as citation networks and previously stored full paper collections, which do not allow systems without that information to offer recommendation services. In order to solve this problem, in this work we propose a framework for finding and recommending research papers that does not use any privileged information. The framework discovers related work through search interfaces available on the Web and recommends the most relevant papers by using only publicly available information. Then, we show how to create and select queries by using information from a single paper, and evaluate for this task the use of different fields from the research paper. We also evaluate the use of different information sources and content-based strategies for paper recommendation.