PRELIMINARY ANALYSIS OF SELF-CONFIDENCE LABELS OF SECOND LANGUAGE UTTERANCES IN AN ONLINE ASSIGNMENT TO DEVELOP A LEARNING SYSTEM WITH FEEDBACK ADAPTATION
Kyushu University (JAPAN)
About this paper:
Conference name: 17th International Technology, Education and Development Conference
Dates: 6-8 March, 2023
Location: Valencia, Spain
Abstract:Learners’ psychological aspects such as anxiety and self-confidence have been seen to affect their second language acquisition (SLA) and Willingness to Communicate (WTC) directly and indirectly. However, it is difficult for teachers to take into account these important aspects since they are different for each student and it takes time. The first candidate to solve this problem is learning systems such as Computer Assisted Language Learning (CALL) systems with adaptive feedback for these aspects. CALL systems provide us with second language (L2) learning environments whenever and wherever even outside the classroom. Some learning systems generate feedback which adapted to the learners from their predicted psychological aspects to facilitate learning, so CALL systems with the adaptation should help the teachers with the consideration.
Among psychological aspects, our focus is self-confidence felt by L2 learners when practicing speaking which is a vital means of communication and anxiety-provoking activity in language class. To our knowledge, no study has focused on self-confidence in L2 speaking even though there are studies to predict learners’ emotions. To predict learners’ confidence, we first need to investigate the relationship between L2 learners and self-confidence in their utterances. It is a step toward developing a learning system for L2 speaking.
In this study, we collected paired data of utterance audio and self-reported self-confidence labels on a 4-point scale. The paired data were collected through our developed prototype system as an online assignment on a whole-semester Japanese course at a university. Fourteen international students participated and labeled their own recorded utterances. In our prototype, they are able to record utterances many times until they are satisfied.
As a result, our prototype has collected about 4,500 paired data including unsatisfied utterances too. We found that it is difficult for a few models to describe each learner because the collected labels are different for each learner. This finding suggests that a predicting model should be flexible for each learner’s confidence. In our next step, we will attempt to estimate self-confidence by utterance data and learning logs.
Keywords: Second Language Learning, Learning system, Speaking, Self-confidence, Web Application.