Reuse of Search Experience for Resource Transformation

The advent of Web 2.0 has created a proliferation of resource sharing sites where individual users tag resources. Retrieval performance is good when users share the same vocabulary, but deteriorates when users have diverging vocabularies. In this paper we propose a novel method of reusing search experience to transform the underlying representation of tagged resources. The aim is to favour those tags that best correspond to community consensus. A CBR approach is presented to learn from user search histories, modifying resource tags in response to implicit user feedback. We evaluate this method on a prototype image retrieval system IFETCH. Our evaluation shows that resource transformation progressively increases the ranking of those images that are generally deemed relevant by similar search sessions. Our results also confirm that the casebase weight update mechanism is more robust to erroneous user feedback compared to a naive constant weight update strategy.