An Extensible Multilingual Corpus of DFA Construction Problems

Abstract

In a country with linguistic diversity, academic instruction may occur in a language in which learners and instructors lack fluency. In such a situation, learners may struggle to answer problems posed by instructors not because they lack the requisite technical skills, but because they are unable to precisely comprehend what is being asked. In this work-in-progress paper, we demonstrate a semi-automated technique to translate a specific category of problems in the undergraduate Computer Science curriculum (the construction of deterministic finite automata) into multiple languages. We show how a corpus of problems (specified in a recently proposed mathematical representation) can be extended in two ways. First, new languages can be targeted by manually translating elements of this mathematical representation into each new language. This is a one-time effort per target language. Second, new problems can be added to the corpus in the mathematical representation at any time, and these are translated into all target languages automatically. We are currently evaluating our technique with English (the language of instruction at our institution) as well as Kannada and Hindi (a majority of our instructors and learners are fluent in at least one of these), and we describe the results of a small-scale (N = 38) study which shows promising results.