This year for the first time, the United States participated in the International Linguistics Olympiad, an annual competition in which students solve linguistics problems, usually in languages they’ve never learned. This year’s Olympiad was held in St. Petersburg, Russia. Adam Hesterberg of Seattle, WA, won the top award in the individual competition.

That’s pretty good for a kid who simply deleted the first couple of email messages he got at the beginning of 2007 about a new computational linguistics competition in the United States. He was no stranger to competitions: in 2003, he was the MathCounts national champ, and this year, he was a United States of America Mathematics Olympiad (USAMO) winner. But computational linguistics was something else: “I’d never studied linguistics,” he said, “and ‘computation’ sounded like boring calculation.”

But when he opened the third message, from Canada/USA Mathcamp, he decided to look into it. It turns out that he was more interested already in linguistics than he’d realized: an information theory course he’d taken at Mathcamp a few years earlier had gotten him thinking about the structure of language. And he’d learned a little about linguistics in his Latin class. He started doing online practice problems for fun. Then sent in his registration forms, and he convinced a few other students at his school to participate as well.

In April, he took the North American Computational Linguistics Olympiad. (See the exam here.) He was in it for fun, but he still managed to place third. Three months later, he left Mathcamp to board a plane to Russia to compete in the ILO.

Since winning the competition, Adam has begun his freshman year at Princeton, where he plans to major in mathematics. He’s on Cogito to take your questions about linguistics and competing.

I enjoyed the problems at namclo.linguistlist.org. Is this what linguists do for a living?

In short, no. Linguists do a variety of things for work, but most of it isn't what's done at the contests. However, something like the contest problems may occasionally come up, and it certainly uses some of the same skills. The situation is similar for math and math contests: the skills they require (except speed) are also important to real mathematicians, but they don't spend their time doing contest-style math.

In the responses below, I distinguish between contest linguistics and research linguistics.

I really like logic. Is linguistics something I would enjoy?

Contest linguistics: almost certainly. They don't (and can't) expect high school students to have any background in linguistics, so contest problems are essentially logic puzzles.

Research linguistics: harder to tell. Try the contests first, and if you enjoy them, take a linguistics class in college.

I don't understand the term “computational linguistics." Where does the computing come in?

The term refers to the use of computationally intensive models to analyze language: the classic problem is machine translation (trying to program a computer to translate text automatically—if you've ever played around with online translation software, you can tell that much work remains here). Another classic example is analysis of authorship: for instance, computer analyses are used to determine whether Shakespeare was secretly someone else and to catch plagiarists.

Computational linguistics doesn't often appear in linguistics problems, for lack of time and computers, but the NACLO/USALO has some—hence the name “North American Computational Linguistics Olympiad" instead of the pattern-following “USA Linguistics Olympiad." On the NACLO/USALO last year (posted on the above site), problem B was definitely computational, and A, E, and F were related.

Which was more satisfying: your success at the USAMO or the ILO?

The ILO, because I did better there than I expected and knew I'd done my best, whereas I know I could have done much better on the USAMO.

Do you think there's a connection between the skills used for linguistics and those used for math? They seem like such different fields, using different parts of the brain.

That a connection exists is easy to demonstrate: 3 of the 8 US ILO team members were mathematicians.

Determining the nature of the connection is harder. Pattern recognition and logic are major areas of overlap. Computational linguistics uses more advanced math, particularly linear algebra and statistics, and occasionally something like quadratic residues are useful—I solved one of the ILO problems this year with their help.

How did you prepare for the linguistics competitions other than doing practice problems? It said you hadn't ever studied linguistics, except just a little bit in a Latin class.

Until I placed third in the NACLO and was invited to the national team, I wasn't serious enough about the contest to do anything just for it—I did all the practice problems I could find because they were fun, and I was competing for the fun of it. After being invited, I attended some online practices with the team and read Crystal's Cambridge Encyclopedia of Language. However, some of the things I'd doing before I heard about linguistics competitions turned out to be excellent preparation: I often get sidetracked into articles on languages or linguistics when reading Wikipedia, and I spent (and still spend) many, many hours on the following thought experiment, which led me to most of the areas of linguistics (some of which might even be applicable to contests, although that's not the point):

You're on a spaceship with a few hundred other people going to colonize a planet in another galaxy. It'll be a long time until you get there—longer than your lifetime—so you decide to make a new and “better" language for the colony, starting from scratch. The colonists will still be human, but they won't have to communicate with anyone using the old languages. What does “better" mean? (I started thinking about this after hearing an estimate of the efficiency of written English (something like 16%)—but you might also want to consider how much ambiguity is acceptable and how hard the language would be to learn). Before you start the language, you'll need to choose its phoneme inventory (the set of sounds a language uses). At some point, you'll need to figure out its grammar, a writing system, vocabulary . . . basically, you'll encounter a lot of fields of linguistics.

What did you enjoy most about the Olympiad?

Meeting my best friend there.

I didn't realize that until several months after the contest. Until then, I'd have said the moment when I realized I could use quadratic residues in solving the hardest problem on the test. (They weren't required, of course, but they made it easier).

What advice would you give someone interested in competing?

Do all the practice problems posted on webscript.princeton.edu/~ahesterb/puzzles.php (not my problems; I'm just hosting my favorite problem site, created by Tom Payne), namclo.linguistlist.org/problems.cfm, and past ILO problems. To actually compete, the contact person (posted on the website) is Dragomir Radev (radev@umich.edu). You'll just need to find a proctor and send in a form.

Aren't there different kinds of linguistics? For speech and for written languages?

There are many fields of linguistics—phonetics, semantics, syntax.... Wikipedia's entry on linguistics has a long list of them.

The article about you said that linguistics was fun, and that's okay, but how can linguistics be used in a useful way?

Like math, there are useful parts of linguistics and parts that seem like they'll never be useful, and I'm more inclined to the former (in both math and linguistics). Nevertheless, a few useful things:

The government (and business) is interested in automatic translation software (e.g. for Russian during the Cold War, and especially Arabic now), which is an active area of research. Search engines need to be able to analyse internet text to sort the information effectively—and, contrary to appearances, not all laws of grammar disappear when people write online. Speech recognition is a popular topic of computer people at the moment.

What was St. Petersburg like?

Interesting. Pretty. More smokers than Seattle or Princeton.

Did you think about majoring in linguistics instead of math after you won?

By the time I'd won (or even knew I'd done well enough to be invited to the team), I'd decided to go to Princeton, which doesn't offer a linguistics major. Ironically, my second choice was MIT, with the strongest linguistics department in the country—but I've stayed sure that I'll major in math, and I'm happy with my choice. I will, however, minor in linguistics.

Learn more about the North American Computational Linguistics Olympiad.