Relevance Feedback in Information Retrieval Systems for Specific Applications

This work is intended to be an analysis and an evaluation about relevance feedback methods in information retrieval. This topic is very interesting because it is discussed since the early seventies and the need for relevance feedback is as big as any time before. Nowadays the amount of information is getting more and more. Therefore the need for advanced information retrieval methods is evident. The work contains a theoretical overview of the relevance feedback methods and how to gain implicitly or explicitly relevance feedback. The theoretical overview forms a good background to give then an insight in the practical work of this thesis where the evaluation framework is described. This contains the requirements, the package structure and the implementation details. In this chapter the three implemented methods - the expansion method, the IDE method and the SVM method - are also described. Then the evaluation itself is discussed. It contains amongst other things the query finding process and the test setting. The evaluation of the methods gives the reader the ability to compare the methods in a practical use case because there a user test is conducted on the three implemented methods. Last but not least a conclusion about the results will be drawn and some insights in further work will be given. The conclusion should give the reader a decision aid which relevance feedback method to use for specific applications based on the results of the evaluation. Additionally a preview to other so-called new or higher sophisticated relevance feedback approaches is given.