Personalized Ontologies for Web Search and Caching Susan Gauch Information and Telecommunications Technology Center Electrical Engineering and Computer Science The University of Kansas Department of Electrical Engineering and Computer Science ITTC Outline Motivation User profiles creation and maintenance evaluation Applications re-ranking (and filtering) search results Web caching Conclusions Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Motivation Decrease access time for Web pages Server approaches - use access logs to decrease access times for popular pages - not tailored to individuals - doesn’t decrease network traffic Network approaches - cache popular pages multiple places in the network - not tailored to individuals Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Personalization Different information needs for different users can we learn user’s interest? - Explicitly? - Implicitly can we use this information? - improved search - improved browsing - faster Web page access Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Intelligent Web Caching Improved (and faster) search results pre-caching all search results expensive - Internet search engines return 50% irrelevant pages improved knowledge of user’s likely behavior - intelligent pre-caching - use past behaviors to predict future behaviors - pre-cache “best” pages close to individuals Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Context ProFusion: www.profusion.com OBIWAN: distributed content based IR Web clustered into regions clustering criteria: content, location, company search: query brokered to “best” regions; within region brokered to most promising sites browsing a region means browsing its sites simultaneously www.ittc.ukases.edu/obiwan Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC User Profiles Applications Usenet news filtering recommendation services: web browsing, books intelligent pre-caching Should accurately reflect actual interests require as little feedback as possible be dynamic Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC User profiles: Creation Obvious and often used: keywords not structured (ambiguous) static have to be explicitly mentioned Our approach watch over a user's shoulder while surfing automatically determine documents’ content central: large ontology (concept hierarchy) Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Document Classification Documents as weighted keyword vectors: n different words -> n dimensions weights based on word frequency and rarity Browsing hierarchy: 10 web pages per node Concatenate them -> keyword vector Content of a page: most similar vector Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Updating profiles Static: document related content: weights of top nodes for surfed document length of page Dynamic: time spent Combine them for instance: weight * (time/length) changes in interest in the five categories User profile: weighted ontology Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Profile evaluation Accordance with actual user interests 10/20 interest categories describe actual interests describe interests “pretty well”: 3.5/5 Convergence stabilization of # of categories over time? do converge after 320 surfed pages! Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Profiles: Summary Stored as weighted ontologies Profiles represent actual interests quite well Up to 150 top categories Two adjustment functions make profiles converge after 320 pages length of page doesn't really matter, but time spent does Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Personalizing Search Results 50% of top 20 results irrelevant Same search mechanism for 200 million people? Goal: identify relevant documents and put them on top of the result list (pre-fetch relevant results) Difficult problem: 10% increase is very good Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Re-Ranking Ranking a function of: search engine's original ranking extents to which top 5 categories describe document's content personal interest in each of these top categories “More relevant items on top of result list”: system’s ability to present all relevant items system’s ability to present only relevant items Professor Susan Gauch December 1999 ITTC Department of Electrical Engineering and Computer Science Recall and Precision Combination: Recall/Precision graphs Example: ranked documents 1,…,20 Professor Susan Gauch 0.6 0.5 precision relevant 2,5,10,14,19 recall points 1/5, 2/5, 3/5, 4/5, 5/5 precisions 1/2, 2/5, 3/10, 4/14, 5/19 0.4 0.3 0.2 0.1 0 0.2 0.4 0.6 0.8 recall December 1999 1 Department of Electrical Engineering and Computer Science ITTC Re-Ranking: Evaluation Overall performance increase of up to 8% at each recall cutoff, up to 10% more relevant documents have been retrieved Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Browsing Assistance Analyze current page locate links Identify which links are most likely to be followed by the user popularity of the link overall relevance of linked page to user’s interests Problem if you have to download the whole page to analyze it, you’ve increased the network utilization Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Privacy Is the user aware that their behavior is being monitored? Can users turn it off? Where are profiles stored? With whom are profiles shared? How are profiles protected? How are profiles used? Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Conclusions Automatic creation of structured user profiles is possible Profiles are reasonably accurate Applications in improving the search quality and Web page access efficiency Evaluation of re-ranking search results: performance increase of up to 8% Professor Susan Gauch December 1999 Department of Electrical Engineering and Computer Science ITTC Future Work Incorporating profile generator into browser Connect system to ProFusion, OBIWAN Personalize structure of ontology Re-train classifier More applications: recommendation service, web caching, browsing, ... Explicit user feedback? Professor Susan Gauch December 1999