Dynamic Web information foraging using self‐interested agents: Application to scientific citations network
Yassine Drias, Samir Kechid · Concurrency and Computation Practice and Experience · 2017
Summary In this paper, a dynamic information foraging approach is proposed. Game theory and more precisely normal‐form is adapted to model the addressed problem. We tackle the issue of Web information foraging as a game played by a set of self‐interested agents that aim at reaching relevant Web pages in a short time. We introduce three kinds of agents along with the strategies followed by each one of them. In fact, the actions performed by the agents would highly depend on the strategy they adopt. We propose a pure strategy, a mixed strategy, and a fully mixed strategy respectively for the three kinds of agents. The agents share a common goal, which is satisfying a specific information need and work together in order to achieve it. The cooperative aspect between the self‐interested agents allows to get information from the Web in an effective and efficient way. Furthermore, we discuss the dynamic aspect of our information foraging approach which is ensured thanks to a modular multi‐agent system architecture. In order to test our Web information foraging system, we performed extensive experiments on theCitation Network Datasetthat includes more than 2.3 million scientific publications. A preprocessing step was conducted with the aim of classifying the publications according to the2012 ACMontology, the purpose being to make our approach scalable. We also compared the results of our approach with two information access approaches. The first one deals with Web information retrieval whereas the second concerns medical information foraging. The experimental results show the superiority of our approach in terms of efficiency and results quality.