Steps Towards Adaptive Intranet Search
Speaker: Dr. Udo Kruschwitz
Affiliation: School of Computer Science and Electronic Engineering, University of Essex.
Steps Towards Adaptive Intranet Search.
Dr. Udo Kruschwitz
School of Computer Science and Electronic Engineering, University of Essex.
Until now most search engines have focused on individual queries and their answers. However, it is increasingly recognised that the key to retrieval is interaction (as evidenced by more and more search options being introduced by the major search engines). This is even more true on intranets and local Web sites where information can be very hard to find. The interaction with the search system can take a variety of forms ranging from the presentation of simple relevance feedback terms to fully fledged dialogues. The problem is that many such paradigms have been suggested in the past and very often they fail because searchers simply don't make use of them. We have performed a number of controlled task-based evaluations on a local intranet that have shown that users strongly prefer a system that makes suggestions for query modifications over a standard search engine. What we now need to find out is how the system performs in the real world. The talk will report on a longitudinal study of our system that has been in place as a search engine on a university intranet for more than a year (resulting in about 700,000 queries so far). We will start with some observations derived from the log files and then give an overview of our ongoing research which aims at learning from past user searches in order to automatically improve the suggestions made by the system.
Udo Kruschwitz is a lecturer in the School of Computer Science and Electronic Engineering at the University of Essex. He holds a diploma in Computer Science from the Humboldt Universitaet Berlin and was awarded a PhD in Computer Science from the University of Essex (2004). His main research interest is the interface between information retrieval and natural language processing. He has successfully led EPSRC and ESRC research projects and is currently investigator on two EPSRC projects and two joint university/industry knowledge transfer partnership (KTP) projects which aim at turning NLP and IR research into practical commercial applications. Udo is the author of the monograph "Intelligent Document Retrieval: Exploiting Markup Structure", published in Springer's Information Retrieval series. He is programme co-chair of next year's European Conference on Information Retrieval (ECIR 2010).