(899.pdf) AUTOMATIC CLASSIFICATION OF TEXT WRITTEN BY EFL LEARNERS BASED ON LINGUISTIC FEATURES AND LEARNER FEATURES

摘要

Automatic classification of text written by learners of English as a foreign language (EFL learners) canrelieve the burden placed on language teachers by reducing the time and effort required to evaluatethe text. In addition, language teachers can use classified text to extract linguistic knowledge that isimportant for EFL learners to know. One approach for automatic classification determines theappropriateness of text by examining linguistic features such as the distribution of adjectives, adverbs,nouns, and verb phrases, as these linguistic features are well-known evaluation criteria. However, thisapproach neglects learner features that indicate how EFL learners wrote the text in terms of the timespent writing and learners' confidence in the appropriateness of the text. These learner features arealso significant indicators of writing proficiency. Writing time indicates the difficulty of writing the textfor EFL learners, as more time is required to write more difficult text. Similarly, confidence is related todifficulty because EFL learners have less confidence in writing more difficult text. Given thisbackground, the present study constructed classification methods for text based on both linguisticfeatures and learner features. These methods were constructed using discriminant analysis. Theexplanatory variables were the linguistic features of the text and the learner features of writing timeand learner confidence. The objective variables were the integrated scores for the grammatical andsemantic correctness of the text. The experimental results showed that the tested methods had higheraccuracy than random chance in leave-one-out (k-fold) cross-validation tests: 71.1% for binaryclassification, 42.2% for five-group classification, and 28.9% for ten-group classification. Thus, weconcluded that these methods are useful for assisting language teachers in assessing theappropriateness of text written by EFL learners.KEnglish as a foreign language, writing proficiency, computer-assisted language learning,automatic classification.