Home // International Journal On Advances in Software, volume 11, numbers 3 and 4, 2018 // View article
Authors:
Jean-Marie Poulin
Alexandre Blondin Massé
Keywords: Lexicons; Learning Strategies; Word Lists; Graph theory.
Abstract:
We examine the structure of dictionaries, more specif- ically the interweaving of links that connect words through their definitions. With few exceptions, all the words used to construct dictionary definitions are defined somewhere else in the dictionary. All these references between words create a network of relations, thus making it possible to use graph theory for the study of dictionary structures. We propose using words learning as an investigative tool. For a given dictionary or lexicon, what would be the best strategy to learn all its headwords? To answer this question, we introduce a formal model and simple graph algorithms. We evaluate several different learning strategies by comparing their learning rate and their efficiency for 8 mono- lingual English-language dictionaries. It turns out that the most significant factor affecting the performance of learning strategies is their ability to break definitions circularity. In other words, the most effective learning strategies are the ones that break definition loops as quickly as possible. We show that a very simple algorithmic strategy, based solely on the vertices out-degree - the number of definitions in which lexemes participate - significantly improves the learning process when compared to psycholinguistic- based strategies. We also put forward that such an approach represents an efficient alternative for the construction of “word lists” used to teach foreign languages.
Pages: 247 to 275
Copyright: Copyright (c) to authors, 2018. Used with permission.
Publication date: December 30, 2018
Published in: journal
ISSN: 1942-2628