Introducing Computational Concepts in a Linguistics Olympiad

Linguistics olympiads, now offered in more than 20 countries, provide secondary-school students a compelling introduction to an unfamiliar field. The North American Computational Linguistics Olympiad (NACLO) includes computational puzzles in addition to purely linguistic ones. This paper explores the computational subject matter we want to convey via NACLO, as well as some of the challenges that arise when adapting problems in computational linguistics to an audience that may have no background in computer science, linguistics, or advanced mathematics. We present a small library of reusable design patterns that have proven useful when composing puzzles appropriate for secondary-school students. 1 What is a Linguistics Olympiad? A linguistics olympiad (LO) (Payne and Derzhanski, 2010) is a puzzle contest for secondary-school students in which contestants compete to solve self-contained linguistics problem sets. LOs have their origin in the Moscow Traditional Olympiad in Linguistics, established in 1965, and have since spread around the world; an international contest http://www.ioling.org has been held yearly since 2003. In an LO, every problem set is self-contained, so no prior experience in linguistics is necessary to compete. In fact, LO contests are fun and rewarding for exactly this reason: by the end of the contest, contestants are managing to read hieroglyphics, conjugate verbs in Swahili, and capable of other amazing feats. Furthermore, they have accomplished this solely through their own analytical abilities and linguistic intuition. Based on our experience going into high schools and presenting our material, this “linguistic” way of thinking about languages almost always comes as a novel surprise to students. They largely think about languages as collections of known facts that you learn in classes and from books, not something that you can dive into and figure out for yourself. This is a hands-on antidote to the common public misconception that linguists are fundamentally polyglots, rather than language scientists, and students come out of the experience having realized that linguistics is a very different field (and hopefully a more compelling one) than they had assumed it to be. 2 Computational Linguistics at the LO Our goal, since starting the North American Computational Linguistics Olympiad (NACLO) in 2007 (Radev et al., 2008), has been to explore how this LO experience can be used to introduce students to computational linguistics. Topics in computational linguistics have been featured before in LOs, occasionally in the Moscow LO and with some regularity in the Bulgarian LO. Our deliberations began with some troubling statistics regarding enrollments in computer science programs (Zweben, 2013). Between 2003 and 2007 enrollments in computer science dropped dramatically. This was attributed in part to the dip in the IT sector, but it also stemmed in part from a perception problem in which teenagers view CS careers as mundane and boring: “I don’t want to be Dilbert,1 sitting in a cubicle programming payroll software my whole life.” This is an unrealistically narrow perception of the kinds of problems computer scientists tackle, and NACLO began in part as a way to publicize to teenagers that many interesting problems can be approached using computational methods. Although enrollments are not yet back to the 2003 levels, there has been a sharp increase since 2007 (Zweben, 2013). The resurgence can be attributed in part to the strength of the IT sector, but also to the realization that computer science is relevant to almost every area of science and technology (Thibodeau, 2013). NACLO aims to be part of this trend by showing students that computer science is used in studying fascinating problems related to human language. Even “traditional” LO puzzles are inherently computational in that they require pattern recognition, abstraction, generalization, and establishing and pruning a solution space. However, we also want to teach computational linguistics more explicitly. NACLO puzzles have featured a wide variety of topics in computational linguistics and computer science; they may focus on the application itself, or on concepts, tools, and algorithms that underlie the applications. Broadly, computational LO topics fall into three types, summarized