Relevance feedback (RF) techniques make use of this fact to automatically modify a query representation based on the documents a user considers to be relevant. RF has proved to be relatively successful at increasing the effectiveness of retrieval systems in certain types of search, and RF techniques have gradually appeared in operational IR systems and even some Web engines. However, the traditional approaches to RF do not consider the behavioural aspects of information seeking. The standard RF algorithms consider only what documents the user has marked as relevant; they do not consider why the user has assessed relevance. For RF to become an effective support to information seeking it is imperative to develop new models of RF that are capable of incorporating why users make relevance assessments.
The underlying assumption of the vast majority of RF theories is that terms occurring more frequently in relevant documents than non-relevant documents tend to be good for retrieving more relevant documents. However, it has been demonstrated in a number of studies that why users mark documents as relevant is as important as which documents they mark relevant, in deciding what further documents to retrieve. This means, in deciding whether a document is likely to be relevant, we not only have to consider which terms are used in documents: we also have to consider how the terms are used in documents.
In this project we view RF as a process of explanation. A RF theory should provide an explanation of why a document is relevant to an information need. Such an explanation can be based on how information is used within documents. We will use abductive logic to provide a framework for an explanation-based account of RF. Abductive logic is specifically designed as a technique for generating explanations of complex events, and has been widely used in a range of diagnostic systems. Such a framework will produce a set of possible explanations for why a user marked a number of documents relevant at the current iteration. These explanations will be based on how information is used within relevant documents. From the set of possible explanations, one explanation, known as the best possible explanation, will be selected to reformulate the query. The choice of the best possible explanation is guided by a number of factors, the main factor being the previous search history.
1. I. Ruthven and M. Lalmas. Selective relevance feedback using term characteristics. CoLIS 3, Proceedings of the Third International Conference on Conceptions of Library and Information Science. Dubrovnik. 1999. ps version pdf version
2. I. Ruthven, M. Lalmas and C.J. van Rijsbergen. Retrieval through explanation: an abductive inference approach to relevance feedback. 10th Irish Conference on Artificial Intelligence & Cognitive Science. Cork. 1999. ps version pdf version
3.I. Ruthven. Abduction, explanation and relevance feedback. Searching for Information: Artificial Intelligence and Information Retrieval Approaches. Glasgow. 1999. Poster paper.ps version pdf version
4.I. Ruthven. Incorporating aspects of information use into relevance feedback. Journal of Information Retrieval. Vol 2. Issue 01. 2000.ps version pdf version
5.I. Ruthven, M. Lalmas and C.J. van Rijsbergen. Empirical investigations on query modification using abductive explanations. ACM SIGIR Conference on Research and Development in Information Retrieval, New Orleans, USA, 2001. to appearps version pdf version
6.I. Ruthven, M. Lalmas and C.J. van Rijsbergen. Combining and selecting characteristics of information use . Journal of the American Society for Information Science and Technology. 2001. to appearps version pdf version
7.I. Ruthven and M. Lalmas Using Dempster-ShaferŐs Theory of Evidence to combine aspects of information use. Journal of Intelligent Information Systems. 2001. to appearps version pdf version