Download Research Tools
How did you do in your high school linguistics courses? What? Your high school didn’t offer any linguistics classes? Well, you’re not alone; it’s difficult to find any secondary school in the world that formally teaches linguistics.
As a scientist, I find that unfortunate. It means that countless bright, inquisitive young people are never exposed to the scientific study of language, a subject that is both immensely important and intellectually satisfying. Analyzing the logical structure of language builds critical skills in computational thinking, a skillset that will be increasingly necessary in today’s digital world. Moreover, linguistics studies make students aware of the diversity of languages and cultures that enrich our world, opening their eyes to the fact that there are many ways to think about things. It also sensitizes them to the plight of dying languages, a process that is accelerating due to the prominence of a few major languages in modern media. Finally, in my country, India, as in many places, the way language is taught in school centers on grammar and literature; consequently, smart, analytically minded students are not exposed to the idea that language can be studied scientifically.
This is why my colleagues in the computational linguistics group at Microsoft Research India and I are staunch supporters of the International Linguistics Olympiad (IOL). This annual event, first held in 2002, is open only to high school students and thus offers a unique opportunity to introduce bright teenagers to the scientific and computational aspects of human language. Every participating country (around 30 in 2014) sends one or two teams of four students each. Participants needn’t have any theoretical background in linguistics. Instead, they solve some of the toughest linguistic problems through sheer analytical abilities and linguistic intuition, as shown in the video below.
Microsoft Research has a rich history in the area of language, having conducted fundamental research on natural language processing, speech-to-speech, and automated translation, so it was only natural that my colleagues and I at Microsoft Research India would see the value of the IOL—but it took a while. India first participated in the IOL in 2009; however, it was only in 2011 that a systematic national selection event, the Panini Linguistics Olympiad was introduced by the University of Mumbai. I came to know about the Olympiad from Prof. Dragomir Radev of the University of Michigan, who is the coach/team leader of the teams from the United States and an active organizer of the North American Computational Linguistics Olympiad (NACLO). He solicited help from Microsoft Research for sponsoring and promoting their Indian counterpart.
The Panini Linguistics Olympiad happens in two rounds: round 1, which involves a written test, takes place across several cities in India. The top 20 to 25 students are invited to round 2, which consists of a weeklong residential camp. During the camp, the students learn some basic linguistic concepts and various problem-solving techniques. Round 2 ends with a test that determines which students will represent India in one of the two teams we send to IOL this year.
2015 participants of the weeklong residential camp
After hearing from Prof. Radev, I scouted out the 2013 Panini Linguistics Olympiad camp. It was a fascinating experience; I enjoyed lecturing and interacting with a bunch of young talented students, as well as designing linguistic puzzles and problems for the selection test. The camp organizers—Prof. Avinash Pandey of Mumbai University and Zarana Sarda, program coordinator for the Panini Linguistics Olympiad—were very motivated; however, the program lacked financial and infrastructural support. There was very little awareness about the Olympiad, which was evident from the fact that only 26 students, mostly from Mumbai, had written the round 1 test that year. So, my colleagues Kalika Bali and A Kumaran and I decided to support the Panini Olympiad with support from Microsoft Research.
In 2014, Kalika and I contacted several linguists and computational linguists across the country and with their help, we set up five regional centers for the Olympiad. We also conducted a promotional workshop for students and teachers in Bangalore. Round 1 participation went up by a factor of 5! The weeklong round 2 camp was organized at Microsoft Research India and we also sponsored the travel of the Indian team to the IOL. I accompanied the team as one of the team leaders.
This year, round 1 took place in January and approximately 150 students wrote the test. Microsoft Research again organized and hosted the round 2 camp, which was held May 24–31, and we will sponsor the trip of one of the Indian teams to the IOL. (Our friends at the Xerox Research Center India are underwriting half the cost of the second team.) Next year, India will host the IOL, which is being hosted in Bulgaria this year.
The Panini Linguistics Olympiad medal winners: juniors in the front row and seniors in the middle row—the seniors will represent India at IOL this year. Tutors and organizers are in the back row.
In short, we are promoting the Linguistics Olympiad in every possible way: by getting more institutes, people, and companies involved; by promoting awareness among students and schools through various events; by funding and organizing the round 2 camp and IOL travel; and by getting involved with the technical activities, such as designing problems, delivering lectures, and coaching the students post-selection. We are already witnessing a very promising pattern: many Indian participants are choosing linguistics for their university studies, and even those who gravitate to other disciplines (mostly computer science) pursue linguistics or languages as a minor.
It is our hope that these efforts enhance the scientific study of language in India and around the world.
—Monojit Choudhury, Researcher, Microsoft Research
What if a mobile game maker could determine the “who, when, and where” that lead to the most satisfying gaming experience? What if they knew precise combination of demographics, location, and recent activities—say young men, in their dorm room, after a meal—that yields the highest satisfaction with their game? Armed with such precise data, the company could push incentives and promote new games to the right people at the right time and place.
This level of data specificity would require granular location data, coupled with mechanisms to match demographic data with dynamic activity detection. Perfecting any one of these components would be a major research project in itself. Putting them together and making them operational in real-world environments would seem a near impossible task. But it’s exactly what researchers at the Singapore Management University’s School of Information Systems are striving for with their LiveLabs Urban Lifestyle Innovation Platform (LiveLabs, for short), a mobile experimentation test-bed deployed across the campus.
According to Rajesh Balan, the principal investigator and co-director of LiveLabs, the goal is to allow in-situ, real-time experimentation of mobile applications and services that require context-specific triggers—all based on real participants using their personal smart phones. As of March 2015, more than 3,000 students had signed up with LiveLabs, and several hundred were active users.
The LiveLabs closed-loop experimentation cycle
The diagram above shows the project’s overall vision. It starts with the collection of sensor and contextual data from a participant’s smart phone (with the participants’ permission, of course). This data is then fused with other data streams, such as location and activity type, to deduce the current user context: determining, for example, that the user is standing outside a store or is with a group of friends. The project team then employs an experiment-creation interface (shown below), in order to determine if the data provides sufficient contextual triggers to warrant sending an experiment, in the form of a customized notification, to the participant’s phone. The investigators then collect data on what the participant actually did after receiving the experiment, using this data to test their hypothesis and improve the system’s efficacy.
Step 1: Participant data is processed in the experiment-creation interface.
Step 2: Experiment is sent to user's phone.
Step 3: Report is compiled after experiment has been run.
Microsoft Research has provided support for LiveLabs, including an award from Microsoft Research Asia to collaborate on indoor location techniques. “The LiveLabs team has a close relationship and long-standing history of collaboration with Microsoft Research,” says Balan. He notes that members of the LiveLabs team regularly visit Microsoft Research labs in Redmond (Washington), Beijing, and India to present their latest research innovations and engage with Microsoft researchers on shared initiatives.
“Many portions of LiveLabs research and deployment—including indoor location tracking, sensing technologies and algorithms, behavioral analytics, and power management—overlap with the agenda of Microsoft researchers,” Balan adds. Jacky Shen of the Wireless and Networking Group at Microsoft Research Asia agrees. “We’re seeing lots of moves in the indoor localization research area. It’s essential for mobile sensing, mobile multimedia, and more general mobile Internet.”
Balan hopes to expand LiveLabs beyond the university campus. “The university campus gives us a good starting point,” he says, but he notes that broader public participation is necessary to adequately test the research theory and model. “We’ve started to look into a large convention center, a commercial airport, or a resort island,” he adds.
Balan is also is looking into further integrating LiveLabs with the Internet of Things, working with Arjmand Samuel from Microsoft Research Redmond. As Balan observed during a recent talk at Microsoft Research Asia, “We can provide access to the LiveLabs experimentation capabilities to any interested Microsoft researchers. In addition, we are very open to further research and professional engagements with Microsoft Research.”
That sentiment runs in both directions, as Microsoft Research is eager to promote collaboration with academics who share our vision of using technology to help solve global challenges.
—Winnie Cui, Senior Research Program Manager, Microsoft Research
Over the years, Microsoft Research Asia has fostered connections among a variety of people in the computer science and technology fields, especially by bringing together Microsoft researchers, interns, and university professors to collaborate on continuing projects. Such was the case when Microsoft Research Asia researcher Xing Xie and Osaka University Associate Professor Takahiro Hara began working together on privacy problems in 2008. Yuki Arase—then a second-year PhD student (under Hara) and Microsoft intern (under Xie)—helped bring the two together. They developed a great team chemistry, which endures to this day, as they continue to work on privacy issues related to location-based services (LBS).
Takahiro Hara, associate professor, Osaka University
LBS privacy protection is an important field for Microsoft Research and the broader computer science community, as we strive to safeguard the privacy of sensitive personal information on users’ electronic devices. The popularity of such LBS as Foursquare and Yelp, which link users to social networks, and Ingress, which provides an augmented-reality gaming platform, poses critical challenges to users’ private information. The security of this information is especially vulnerable as users use mobile devices to access cloud-based services.
Many past studies on location privacy preservation were based on unrealistic expectations about the user's movement pattern, such as assuming it is known in advance or that it will follow a very simple model. Professor Hara and his collaborators have taken a more advanced approach, in which they design and deploy a protocol based on the movements of “dummy” protocol applications that follow more realistic user mobility patterns.
This method can be readily deployed on any LBS that accepts multiple requests from users, eliminating the need for a third party to preserve privacy. Nor does it require changing server programs, because they are expected to run on LBS client-side applications. And since this approach assumes a more realistic model of mobility, it can be deployed in a wider variety of LBS types.
Currently, the collaborators are working to optimize their approach in two ways. The first step is to make reactive, real-time changes in the dummy’s movements, so that they mimic the user’s mobility pattern. The second step is to generate a dummy that takes into account the user’s preferences for visiting various points of interest.
The collaborative approach works not only in research but also in education. So, for example, undergraduates at Osaka University are producing dummy protocols in collaboration with Microsoft Research. The class has workshops of three or four students who team up and log their GPS locations from smart phones when walking though campus. Students feed this logged data into the LBS-privacy application and verify the difficulty of discerning the original user’s movements from the dummy locations produced by the protocol. Students praise how this real-life application has helped them understand the concept of privacy preservation.
Screen shot of dummy-protocol application
Through the efforts of students, post-graduate interns, researchers, and professors, Microsoft Research and its academic partners help advance computer science and technology, and help supply society with much-needed, talented engineers. For example, over the past seven years, five PhD students from Professor Hara’s labs have participated in Microsoft Research internships in web mining, multimedia, and information retrieval systems. These interns spent three to ten months in Microsoft Research Asia, conducting intensive research projects under the supervision and mentorship of various Microsoft researchers, all while experiencing a different culture and forging international friendships After graduation, Microsoft interns have joined respected companies, such as Fujitsu and KDDI, as engineers and researchers, putting their collaborative training to work for society.
The professional and personal relationships created during Microsoft Research’s collaborative endeavors often influence the participants well beyond the timespan of the specific project. As Professor Shojiro Nishio, a former Osaka University vice president and director of the university’s Cybermedia Center observers, “Our long-time collaboration with Microsoft Research has made significant contributions for not only research outcomes in top-ranked journals and conferences, but also in the education of young students, many of whom are now internationally leading researchers.”
This brings us back to Yuki Arase, the student who instigated the collaborative LBS-privacy project. After working for four years as a researcher in the Natural Language Computing Group at Microsoft Research Asia, she is now an associate professor at Osaka University’s Graduate School of Information Science and Technology and is co-leading a lab on big data engineering, educating the next generation of ambitious graduate students.
Professor Arase presenting at Microsoft Research Korea • Japan Academic Day
Professor Arase is thus a prime example of the fruitful chain reaction of collaboration. Her growing experience and expertise have produced continued achievements, which is precisely what we hope for from every Microsoft Research intern. As we strive to broaden the interaction between academia and Microsoft Research, we anticipate an exponential growth of such positive chain reactions resulting from our academic collaborations.
—Sean Kuno, Research Program Manager, Microsoft Research