Identification of Low-engaged Learners in Robot-led Second Language Conversations with Adults

The main aim of this study is to investigate if verbal, vocal, and facial information can be used identify low-engaged second language learners in robot-led conversation practice. experiments were performed on voice recordings video data from 50 conversations, which a robotic head talks with pairs adult using four different interaction strategies varying robot-learner focus initiative. It was found that these robot influenced learner activity engagement. verbal analysis indicated low rated the significantly lower two out scales related social competence. acoustic vocal video-based analysis, based manual annotations or machine learning classification, both showed engagement robot’s competencies consistently, several cases significantly, lower, addition effectiveness lower. agreement between automatic identification face videos further adequate for future use. These constitute first step towards enabling adaption learners’ through within- between-strategy changes learners.