Zhai Inducted into Inaugural Class of New ACM Information Retrieval Academy

Illinois CS professor ChengXiang Zhai was one of 25 individuals recently inducted into the first class of the ACM Special Interest Group on Information Retrieval (SIGIR) Academy, which honors researchers who have made significant, cumulative contributions to the development of the field of information retrieval (IR), the science behind all search engine applications.

A prominent IR research leader, Zhai is well known for developing intelligent information systems to help people manage and draw insights from large amounts of text data. The robust models, algorithms, and tools he creates can be used to enhance decision making across a variety of fields, including medicine, online education systems, business intelligence, and security.

“I’m very honored to be recognized with this elite group of outstanding scientists,” said Zhai, a Donald Biggar Willett Professor in Engineering. “This recognition is for all my students and collaborators in academia and industry, as well, and it inspires me to tackle new challenges facing the creation of systems that help humans optimize complex decision making.”

Zhai is known for introducing a theoretical foundation for using language models in IR. The accuracy of an IR system is mostly determined by the retrieval models used in the system.  Zhai’s Dirichlet prior smoothing method is now the most popular smoothing method widely used in IR applications and has been implemented in all the major IR toolkits, including Lucene (used by many companies).  He also developed a novel axiomatic retrieval framework, which for the first time enabled analytical prediction of the performance of a retrieval model. 

In his most recent work, Zhai is developing a theoretical framework that mathematically describes the interaction between a user and an intelligent system (search engine or chatbot, for example) in order to produce better and more relevant results via human-machine collaboration. His cooperative game-playing model involves two players—the computer system and the user; the objective is to help the user finish a task with minimal effort.

The game starts when a user types in a query. The computer responds by creating an interface card, through which it can interact with the user in many ways such as showing search results, asking a clarifying question, or providing a link or visualization. The game goes on from there as the user solicits the information needed from the computer and provides feedback.

“We want to mathematically describe this kind of interaction and optimize a sequence of decisions over the horizon,” said Zhai.

In order for the system to optimize this kind of game, the system has to have a notion of what the user will do at each move of the game. To achieve that, Zhai is building simulators of users to mimic how an individual user will operate. He also designed algorithms that allow the user to browse adaptively so results are viewed on screens of different sizes.

Zhai has a long track record of fruitful industry interactions with companies such as Microsoft, Google, Yahoo, HP, and IBM.  He currently serves as associate director of the IBM-Illinois Center for Cognitive Computing Systems Research (C3SR), a multi-year joint venture between IBM Research and University of Illinois at Urbana-Champaign to solve the most pressing issues facing the new computing era of AI and cognitive computing. As part of that, Zhai is conducting applied AI research and building general interactive intelligent systems to leverage massive amounts of data and hybrid cloud computing for optimization of complex decision making in many application domains.

Like many other 2021 SIGIR Academy inaugural inductees, Zhai is a fellow of ACM. Earlier in his career, he received the prestigious Presidential Early Career Award for Scientists and Engineers (PECASE) and NSF CAREER awards.