Investigation of Multiple Recognitions Used for EFL Writing in Authentic Contexts.

AbstractRecognition technologies had been prevailing and widely used for EFL learning. We investigated the different recognitions used for EFL writing based on image-to-text, translated speech-to-text, and location-to-text recognitions – ITR, TSTR, and LTR. A quasi-experiment was implemented for 12 weeks in a vocational high school with experimental and control groups in two stages. Pre-test, posttests 1 and 2, questionnaires, and interviews were conducted and analyzed. Experimental learners, who wrote writing based on ITR and TSTR, outperformed control learners who wrote that based on TSTR only. Also, the experimental learners, who wore writing based on ITR, TSTR, and LTR, outperformed the control learners who wrote that based on ITR and TSTR. Particularly, LTR was beneficial for identifying controlling ideas and addressing the writing topics. ITR was beneficial for brainstorming and generating more ideas. TSTR was beneficial for yielding and transferring writing contents into words. The multiple recognitions were beneficial for most EFL writers, especially for low-ability language writers. Most writers were interested in describing based on authentic context learning. However, they complained about the low accuracy of LTR and TSTR and the difficulty of ITR texts when writing. Accordingly, the LTR database with various categories of places, the generation of ITR based on the language abilities of learners, and the higher accuracy of TSTR should be strictly considered when applying multiple recognitions for EFL writing. KeywordsEFL writingRecognition technologyMultimedia learningAdaptive technologies