Purchasing decisions can be hard to make. For example, the decision to purchase a house by a person can be based on many parameters such as the price of the house, distance from her workplace, quietness of the neighborhood, etc. It is hard to capture all such parameters in a recommender system or a knowledge-sharing website. Further, many such parameters are highly subjective and difficult to evaluate or quantitatively compare with respect to other parameters. Therefore, people very often refer to their friends for advice, or search for relevant blogs on the Internet that may reveal some new insights. However, how can the person be sure that opinions from friends represent fairly diverse viewpoints so that she can make a wise choice? How can she judge the reliability of information obtained from blogs? An attempt has been made to address such questions in this work through the use of social networks. A design for an infrastructure for information recommendation is proposed, which ensures that the information recommended to people is relevant, diverse, and reliable. To the best of my knowledge, I am not aware of any other recommender system that considers all these three aspects of information. Most recommender systems address the need for relevant information through personalization, or reliability through reputation systems, but diversity is often absent in these systems. The same concerns for relevant, diverse, and reliable information surface not just in reference to e-commerce related information, but also for political news, or news that guide personal and professional economic decisions. Therefore, rather than considering only e-commerce related recommender systems, the problem is considered in a much broader context of recommender systems for participatory media.
Subjects: 1.10 Information Retrieval; 7.1 Multi-Agent Systems
Submitted: May 6, 2007