Adapting meta information retrieval to user preferences and document features

Human and contextual factors of information processes need to be integrated into Information Retrieval (IR) models, although often ignored. This article introduces an approach to adapting of an IR system by optimizing the choice of retrieval functions based on user preferences and document features. The framework is provided by MIMOR (Multiple Indexing and Method-Object Relations), a novel approach in Information Retrieval which exploits users’ relevance feedback information to model method-object relations. By processing the knowledge from feedback, MIMOR automatically chooses the optimal retrieval methods for a specific situation. In this paper, MIMOR is presented and a formal model introduced. In the last section, the model is extended to incorporate user preferences in order to allow users to fine tune his/her own Meta IR System. All users contribute to a public model providing relevance feedback information. They are also to control the weighting of the private versus the public model during retrieval.