A Contextual Computing approach towards Personalized Search Jim Pitkow August 2002
Contextual Computing The enhancement of a user’s interactions by understanding: the user, their context, and the applications/information being used, across a wide set of user goals. Actively adapting the computational environment, for each and every user, at each point of computation Not just about modeling the user’s preferences and behavior or embedding computation everywhere
Evolution: Information Retrieval Methods of Applying Relevance Multidimensional Inferencing User Model Based + Link Analysis Popularity (Google, search engines after 1999/2000) (DirectHit, most search engines after 1999/2000) Popularity Based + Boolean Vector Space (Library and legal systems, Lexis-Nexis, West Law) (Verity, search engine companies till 98) Content Based Accessibility of Information/Volume
Personal Search Architecture Search Enterprise Outride Personalized Search System Query Augmentation User Query Search: Web Demographics Click Stream Commerce: Retail Result Processing Result Set Search History Retrieval over new spaces (Excite’s) Provide retrieval from any Web spaces Contextualization Leverage the customer’s current behavior to make search and targeted Excite alerts smarter and more relevant in real-time Personalization Leverage the customer’s past preferences to individualize results Application Usage Commerce: B2B Contextualized Client Interface Outride Schema User x Content x History x Demographics Search Engine Schema Keyword by Doc IDs by Link Rank
End-User Benefits: Be More Productive SOURCE: ZDLabs/eTesting, Inc. October 2000 % Increase from Outride Enabled Search 130.2% more time 93.7% more time 107.9% more time 114.5% more time Search Engine Craig Time in Seconds
Novice versus Expert Gains Average Time to Complete Task Craig User Skill Level