Jian-Yun Nie's career has taken him from China to France and finally Montreal, where he specializes at UdeM in information research.
His Université de Montreal office is in the André-Aisenstadt building, at the end of a hall on the second floor. Professor Jian-Yun Nie goes to work there every day, looking for ways to improve how search engines work. Nie specializes in information research. Like his small office, his area of study has evolved on the margins of the larger world of databases and artificial intelligence. Nevertheless, it is vital work. Everyone uses search engines.
Back in the 1980s, when he started out, a young researcher had to have a fair bit of gumption to choose computer science as an academic discipline. “At the time, working in information research was a lot like working in a data centre: all you had at your disposal was a text-based database to llok things up in," Nie recalls. "Another standard application back then was to automate library searches. Today, our work goes way beyond those boundaries."
Of course, the game-changer in this and many other fields was the arrival of the World Wide Web. To this day, new data emerge and are generated through interaction between users, and they help improve the quality of research. That process of improvement has been the focus of Nie’s work for almost 30 years now. Despite the progress made, however, he's far from satisfied with the results.
When you submit a query in a search engine, he explains, you don’t necessarily find what you’re looking for, which is at once frustrating and stimulating. “It tells us that the system doesn’t understand everything and that there is still a lot to be done," he says. "And that’s what keeps me going.”
Microsoft is a partner
Nie is working more and more with top technology companies like Microsoft to figure it all out. Partnerships like these give him access to real data from search engines so he can tackle concrete issues. “New problems often arise in companies that work in direct contact with users," he points out. "Universities don’t have that kind of contact, so we can’t know how users search the web."
Every query entered into a search engine is registered. The same goes for every result a user clicks on and opens. Those data allow researchers to observe how users interact with the engine, and to judge the quality of the results. The findings serve to help refine search algorithms and improve results next time a similar query is processed.
More collaboration between industry and university researchers is in everyone’s interest, Nie believes. Microsoft, for instance, optimizes its tools as it advances research in the field by partnering with academic researchers. “In many cases, companies have in-house expertise, but they also benefit from long-standing relationships with external colleagues.”
Montreal via France
Nie first took an interest in information research in 1985, when he left his native China to begin doctoral work at Université Joseph-Fourier in Grenoble, France. “At the time, very few Chinese students left home to study abroad," he recalls. "I considered myself very lucky. In all, 120 of us left that year to study in France. Grenoble was a well-known centre for IT. My principal in China had been there and was impressed with the level of expertise.” After a few months of intensive language study, Nie was even able to communicate in French.
After completing his degree, the young researcher made the decision to come to Montreal. He knew he wanted to pursue a research career in search engines and AI. The Montreal Institute for Learning Algorithms (MILA) didn't exist yet but UdeM did have a well-established Computer Science and Operations Research department, with automated translation among its top disciplines.
“I worked in information research and languages, so I was somewhat familiar with machine translation," he says. "That’s how I came to hear about Université de Montréal, which had a very good reputation thanks to TAUM, its original machine-translation system.” In 1991, Nie started work INCOGNITO, the university's INCOGNITO computing lab. Re-named the RALI Lab in 1987. it specializes in natural language processing and its applications.
Since then, information research has come a long way, but researchers in and around Montreal remain scarce. “There aren’t a lot of us, but I do have colleagues who are working on automatic processing of natural languages," Nie says. "Since information research is so strongly linked to the processing of natural languages, there are a number of collaborative projects going on, so I don’t feel too lonely.” Montreal is well-represented in the global community of information researchers: in 2011, Nie served as president of the annual conference of the Association for Computing Machinery's Special Interest Group on Information Retrieval (SIGIR), and has been asked back to head up the program committee for the organization's next conference, in 2019.
The next level
Work on AI is starting to influence information research and those working in it. While it’s not yet widespread, researchers like Nie see a great deal of potential in pushing the field to the next level: the study of meaning. Every effort made to explore meaning in search-engine processing is worthwhile, they believe.
Search engines have yet to reach the same level of comprehension that two people holding a conversation have. “We don’t exactly understand the meaning of a text or a query,” Nie explains. “The system recognizes words, but we’re still a long way from representing meaning and reasoning.” And that’s the main challenge for most natural-language processing applications, he believes. "When you’re looking for information, you have to understand what a text means and what you’re looking for. That’s what allows us to make comparisons.”
Fortunately, he has a lot of students interested in the challenges facing information research. Besides search engines, their contributions can be found in a wide variety of applications – proof, if any is needed, that there's a bright future in the pursuit of knowledge.
(Article by Catherine Mathys)
Université de Montréal
Tel: 514 343-7593