Abstract:

In this study, we focus on particle errors and discuss an annotation scheme for Korean learner corpora that can be used to extract heuristic patterns of particle errors efficiently. We investigate different properties of particle errors so that they can be later used to identify learner errors automatically, and we provide resourceful annotation guidelines. We present issues that are relevant to learner error annotation including how to classify particle error types, present correct tokens, mark overlapping error types, and so on. Accurate annotation of particle errors will be useful for extracting relevant error rules and will provide substantial benefits to the development of ICALL systems. We argue that it is necessary to link annotation with feedback procedures to increase the efficiency of systems. Furthermore, we investigate whether heritage learners and nonheritage learners generate different error patterns and discuss significant implications for heritage versus nonheritage language learning.

Author: Sun-Hee Lee, Seok Bae Jang, Sang-Kyu Seo