Assessing the AWE-Based Teacher-Assisted Feedback Model for College English Writing Teaching at the Application-Oriented University

Abstract

The Ministry of Education in China promotes the full use of high-quality teaching software and teaching resources to deepen the reform of college English teaching for establishing new teaching models. Scholars generally agree that the automated writing evaluation (AWE) is feasible in English writing teaching; however, the auto-recognition of error categories needs to be improved. Previous studies have mostly focused on its application and reliability, and few have examined the AWE-based teacher-assisted feedback mechanism. We aim to examine and evaluate the feasibility and effectiveness of the mechanism by the case study of an AWE platform in China via a teaching experiment, corpus extraction from misjudged errors and its statistical analysis. The specific research questions to be addressed are: 1) What types of AWE misjudgments are actually manifested, and 2) Can the AWE-based teacher-assisted feedback model compensate for its shortcomings? It argues that under the AWE-based teacher-assisted feedback mechanism, 1) misjudgments of complex semantic recognition, complex sentence recognition and the auto-recognition of content words are likely to be commonly-shared features of the AWE platforms used in application-oriented universities in China, thus providing a general direction for operating the targeted teacher-assisted feedback mechanism; 2) the submitted times and the scoring indicates a linear correlation between the number of students’ submitted times and their final scoring, and this correlation is particularly evident for individual students; when the EFL learners received the writing program under the integrated feedback mechanism, their writing and translation increased accordingly by encouraging multiple submissions, scoring refreshments, multiple feedback by means of online, offline and teacher’s artificial involvement; 3) the AWE still needs to be optimized with respect to the scale of corpus data, multiple error correction algorithms, i.e., the grammatical error correction algorithms based on language models, machine translations and grammar rules.