Dictionary-based MWE Level List and its Application

Abstract

Chung, Woojin, Cho, Hyowon Cho, & Song, Sanghoun. 2020. Dictionary-based MWE Level List and its Application. Korean Journal of Linguistics, 45-3, 671-701. In recent years, deep learning research has required a huge resurgence of interests in the domain of MultiWord Expressions (MWEs). Neural networks now reached a sufficiently high level of performance and MWEs, which is a notably unique and important unit, should be examined to improve the performance of the model. In the scope of our research, we have engineered a MWE list that various NLP tasks could leverage. While some MWE lists are already available, such as the phrasal expression list and PHaVe list, they only provide a narrow range of MWEs and are intended for a receptive purpose. Our MWE list contains a wide range of MWEs with the level of difficulty of each MWE. It covers almost every MWE that native speakers know of. In addition, we constructed the list of general-purpose by applying the unit of lemma. Performance of MWE processing will be improved as we used refined dictionary data and considered direct object of phrasal verbs. The development and rationale of the list will be discussed below and the analysis that measures the quantity of MWEs in each level are in one-stop and IELTS corpus. Our list will increase the MWE processing capacity for NLP tasks as well as pedagogical usage by education for L2 learners. (Korea University)