Design and evaluation of an intelligent program for analyzing writing errors: Korean EFL students in junior high school
Jie-Young Kim, Teachers College, Columbia University, United States
Teachers College, Columbia University . Awarded
In this dissertation, a prototype error-analysis program, called the "Intelligent Computer-Assisted Writing Tool" (ICAnWRIT) was developed for junior high school students in Korea, who were learning to write in English. Many commercial grammar checker programs are available, but they are not suitable for EFL classes because they are not equipped with a parser and thus cannot accurately analyze students' erroneous sentences.
It has been suggested that commercial grammar checkers could be improved by (1) basing them on an error analysis of writing by a target audience and (2) limiting their subject domain. Based on these suggestions, writing samples were gathered for error analysis from 200 tenth grade Korean students who wrote an essay about themselves and their family. The error analysis revealed common types of errors which provided the basis for developing ICAnWRIT.
ICAnWRIT was equipped with (1) a parser consisting of an ATN grammar, which incorporated syntactic structures appropriate to the students' level of English; (2) a lexicon based on vocabulary specified in the curriculum and found in the students' writing; (3) and a parsing algorithm in which top-down, nondeterministic and backtracking parsing strategies were employed. A student modeling technique based on intelligent tutoring systems was applied to the system by (1) adding lexical items that students tended to use incorrectly; (2) incorporating pattern-matching procedures for incorrect expressions that the students used frequently; and (3) adding erroneous paths for incorrect sentences that students frequently used.
Upon completion of the system, an evaluation of the program was performed. By using a second corpus of 100 student-writing samples, the accuracy rate for detecting errors by ICAnWRIT was compared with that of a commercial program used extensively in Korea. Z-test values were obtained to test the significance of the differences between the two programs. The results of the evaluation showed that ICAnWRIT was significantly more accurate than the commercial program. The evaluation also suggested that ICAnWRIT could be improved by expanding the lexicon and the grammar and by implementing semantic interpretations developed in computational linguistics.
Kim, J.Y. Design and evaluation of an intelligent program for analyzing writing errors: Korean EFL students in junior high school. Ph.D. thesis, Teachers College, Columbia University. Retrieved March 26, 2019 from https://www.learntechlib.org/p/126129/.
Citation reproduced with permission of ProQuest LLC.
For copies of dissertations and theses: (800) 521-0600/(734) 761-4700 or https://dissexpress.umi.com