1/5/2010 Kymberly Burkhead-Dalton
Written by Kymberly Burkhead-Dalton
University of Illinois computer science professor ChengXiang Zhai was named a 2009 recipient of the ACM Distinguished Scientist Award. The distinguished award recognizes individual contributions to both the practical and theoretical aspects of computing and information technology, and is given to ACM members with at least 15 years of professional experience who have achieved significant accomplishments or have made a significant impact on the computing field.
Professor Zhai ‘s research spans several related fields including information retrieval, natural language processing, machine learning, data mining, and bioinformatics.
His primary research interest is developing techniques for managing and exploiting large amounts of text information, such as news articles, email messages, scientific literature, government documents, and all kinds of Web pages.
“With the dramatic growth of online information, we are overwhelmed with huge amounts of information and have an urgent need for powerful software tools to help manage and make use of it,” explains Zhai. “I work on a variety of general techniques for searching, filtering, organizing, and mining text information and develop applications in multiple domains including Web, email, and literature.”
Zhai and his team are tackling the problem from a variety of angles, from personalized search, to text mining and information retrieval solutions that better enable task support and decision making capabilities.
Personalized Search Solutions
One of the projects that Zhai is working on is the User-Centered Adaptive Information Retrieval (UCAIR) system, which currently works with the Yahoo! search engine. The software is able to learn what you are interested in and pushes out the recommended information the next time you search.
“UCAIR naturally optimizes search results for you without requiring extra effort on users. It learns your information need based on your search history and through observing how you search”, says Zhai.
“For instance, if a user types in the word jaguar in a query, the search results are going to contain mixed results with Jaguar cars and Jaguar cats. If the user clicks on the car search results, the software saves the information and adapts immediately to the user’s interest by pushing away the animal search results as the user continues to view more search results ”, says Zhai.
A significant advantage of the UCAIR system over other personalized search systems such as Google’s personalized search is that the system helps protecting a user’s privacy by sitting on the user’s own computer. The UCAIR demo system is available at http://sifaka.cs.uiuc.edu/ir/proj/ucair/download.html.
Professor Zhai currently partners with Surf Canyon, a technology company that uses semantic real-time implicit personalization to improve search result relevancy. The technology, developed initially in Zhai’s lab, works by re-ranking search results “on the fly” based on a user model generated from real-time selections. The primary goal is to use more user information to assist the user in finding relevant information that's buried within the often overwhelming amount of search results. Zhai and his students continue to work with Surf Canyon to extend the technology and develop new features.
The long-term vision of the UCAIR project is to develop an intelligent personal search agent that sits on the client-side, integrates information around a user, and provides highly personalized and optimized information service to people, which can eventually go beyond search toward task support.
Collaborative and Social Surfing
Another project Zhai is working on is collaborative and social surfing with a multi-resolution topic map. The current search engines support querying well with only limited support for browsing. However, when a user can’t formulate effective queries as often happens in exploratory search, browsing can be very useful. Zhai and his students have developed techniques to organize search logs into a large-scale multi-resolution topic map to guide a user in browsing into relevant information without needing to formulate a query in the same way as a geographic map is used to guide people in finding attractions when touring a city. To view a demo about browsing with a Topic Map go to http://sifaka.cs.uiuc.edu/ir/proj/csurf/.
“Both the queries entered and the web pages viewed by users can be regarded as information foot prints left by users in surfing the information space. The topic map organizes these foot prints as a map so future users can follow previous users’ foot prints. More users means more effective browsing, achieving the effect of collaborative surfing,” says Zhai.
When combined with a social network, such a topic map can also support social surfing where users can potentially see and follow the footprints of their friends. Combining search and social networks is a major direction that Zhai and his students are currently working on, aiming at connecting users with the right information at the right time. As an example of this work, they have recently developed a news recommender application on Facebook. Facebook users register a community by providing a keyword description and a set of news sources. The system then fetches the news articles and filters them based on the community description to prepare daily news digest. To view the demo and learn more about the Facebook project go to http://sifaka.cs.uiuc.edu/ir/proj/rec/.
Aiding Decisions
Search is a means, not an end; people generally search information to perform a task or make a decision. Zhai’s work in text data mining is aimed at helping users harness information in order to make decisions. Data mining is already used on websites such as eBay and Amazon to help users find products to purchase However, it is still difficult for users to digest the large amounts of comments made about products or sellers by customers. Zhai and his students have been working on general text mining techniques that can integrate, summarize, and analyze large amounts of opinions about a topic expressed in scattered blog articles, product reviews, and forum posts to help people make decisions.
For example, their opinion integration techniques can integrate scattered opinions about a product such as iPhone in blog articles with a well-formed product review of iPhone in CNet to generate an integrated opinion summary for iPhone where representative scattered opinions would be aligned to the well-written review along meaningful aspects such as battery life, activation, and wireless connection. Their sentiment summarization techniques can generate a tabular summary of positive and negative opinions in different aspectual dimensions of a topic as well as trends of these opinions over time. Recently, they have also developed a way to generate a contrastive summary of mixed and contradictory opinions in which comparable positive and negative opinions would be aligned so that a user can easily understand why some people hold positive and others hold negative opinions about a topic. A demo of this technique can be found at: http://sifaka.cs.uiuc.edu/ir/data/cos.
Ideally, in the long run, Professor Zhai wants all these projects to work together to form an intelligent text information access and analysis system that would know well about a user’s information need, provide flexible multi-mode information access, and go beyond search to help users finish all kinds of information-dependent tasks.