Overview of UCAIR (User-Centered Adaptive Information Retrieval) Project

 Zhai and UCAIR in the News IR Group, UIUC  

A common major limitation of existing retrieval models and systems is that the retrieval decision is, in general, based solely on the query and document collection; information about the actual user and the search context is largely ignored. This limitation makes the retrieval performance of existing IR systems inherently non-optimal, as seen clearly in the following two cases:

• Different users may use exactly the same query to search for different information, but existing IR systems return the same results for these users. For example, the query ``IR applications'' on Google returns a mixture of documents about ``information retrieval'' applications and ``infrared'' applications, as ``IR'' can be an acronym for both information retrieval and infrared. Without considering the actual user it is inherently impossible to know which sense ``IR'' refers to.
• A user's information needs may change over time. The same user may sometimes use ``java'' to mean the Java island and some other times use ``java'' to mean the programming language. Without recognizing the search context, it would be again inherently impossible to recognize the correct sense.

The UCAIR ( pronounced as "you care") project seeks to break this limitation of the existing retrieval methods and formally develop a new retrieval paradigm called user-centered adaptive information retrieval (UCAIR), in which user information and search context are both exploited to improve retrieval performance. A lot of user information and search context can be used to improve search accuracy. For example, knowing that "infrared" never occurred in any document that the user has browsed in the past 3 months but "information retrieval" occurred many times would strongly suggest that "IR" in the user's query means "information retrieval". The use of such user information allows a system to customize the search results according to the user. Another example is, when a user submits the query "Java", if the search system knows that the previous query submitted by the same user is "Indonesia travel", it can infer that the user is more likely interested in tour information about the Java Island instead of the Java programming language. The use of such information about the search context allows a system to adapt to the dynamic change of a user's information needs.

The research of UCAIR includes (1) developing a new UCAIR framework based on Bayesian decision theory; (2) developing new language models to exploit user information and search context to improve retrieval accuracy; (3) developing new retrieval methods to optimize the long-term retrieval utility over an entire retrieval session; (4) developing new retrieval methods to leverage user similarities to better infer one particular user's information need based on information about other similar users; and (5) developing prototype UCAIR systems for searching the Web and bioinformatics literature.

The UCAIR project is supported by the National Science Foundation via a CAREER grant IIS-0347933