A Study of Self-Organizing Map in Interactive Relevance Feedback

With the vast amount of potential relevant documents on the Web, a key question for a retrieval system is how to achieve a high accuracy retrieval under Web setting. The work reported in this paper is our latest effort in examining the effectiveness of applying self organizing maps (SOM) in clustering base interactive relevance feedback. Our experiment results demonstrate that SOM can be used as a clustering tool to generate terms for interactive relevance feedback, which produced a significant improvement over the baseline when measured by R-precision. Our results also show that a preferred strategy for combining blind relevance feedback with interactive relevance feedback is to perform the former first then the latter