230 likes | 324 Views
Personalized Search. Xiao Liu xl2230@columbia.edu. Background. My presentation will be based on my paper “Analysis and Evaluation of Personalized Search Technologies”. Query Words. Ranked List. User Context. Query Words. Ranked List. Domain Context. Task/Use Context.
E N D
Personalized Search Xiao Liu xl2230@columbia.edu
Background My presentation will be based on my paper “Analysis and Evaluation of Personalized Search Technologies”.
Query Words Ranked List User Context Query Words Ranked List Domain Context Task/Use Context What’s Personalized Search?
Personalization and Search • Source of personalization • How to get personalized information? • User modeling in personalized systems • How can we model a person’s interests? • Three types to implement personalized search • What are the main features for these types? • Comparison between explicit and implicit ways • What are the pros and cons for each type?
Personalization and Search • Sources of personalization • User data: content-based • Choose right categories • Mark the relevant documents • Usage data: behavior-based • Click-through • Selecting a particular article • User Modeling in Personalized Systems • Three types to implement personalized search • Comparison between explicit and implicit ways
Profile Information • Server information • Web page index • Link graph • Group behavior • Behavior-based • Click-through • Selecting an article • Content-based • Choose right categories • Mark relevant documents
Server-Side v. Client-Side Profile • Server-side • Pros: Access to rich Web/group information • Cons: Personal data stored by someone else • Client-side • Pros: Privacy • Cons: Need to approximate Web statistics • Hybrid solutions • Server sends necessary Web statistics • Client sends some profile information to server
Overview • Sources of personalization • User modeling in personalized systems • In retrieval process • Re-ranking • Query modification • Three types to implement personalized search • Comparison between explicit and implicit ways
Overview • Sources of personalization • User Modeling in Personalized Systems • Three types to implement personalized search • Explicit feedback personalization • Implicit feedback personalization • Combined feedback personalization • Comparison between explicit and implicit ways
Explicit feedback personalization • Adaptive Result Clustering • Needs external feedback • Users’ additional effort are always involved • Supports the reuse of clustering
Web search engine - KARTOO Organizes the returned resources on a graphic interactive map The size of the icons corresponds to the relevance of the site to the given query Closed down in January 2010
Implicit feedback personalization • Without requiring any effort from the user • Based on the user’s profile and prior behavior • Current Context • Search Histories
Combined feedback personalization • Collaborative Search Engines • An emerging trend for Web Search
Overview • Sources of personalization • User Modeling in Personalized Systems • Three types to implement personalized search • Comparison between explicit and implicit ways • Pros and cons of explicit feedback
Explicitly vs. Implicitly Query Words columbia university sportswear NYC or British? • Explicit • User shares more about query intent • User shares more about interests • Hard to express interests explicitly
Learning More Explicitly v. Implicitly • Explicit • User shares more about query intent • User shares more about interests • Hard to express interests explicitly • Implicit • Query context inferred • Profile inferred about the user • Less accurate, needs lots of data
Summary • Source of personalization • User data and usage data • User modeling in personalized systems • In retrieval process, re-ranking and query modification • Three types to implement personalized search • Explicit, implicit and combined • Comparison between explicit and implicit ways • Collaborative search is an emerging trend