Towards computer-assisted language learning with robots, wikipedia and CogInfoCom

Abstract

Computer-Assisted Language Learning is being transformed by the use of humanoid robots combined with Wikipedia-based information access and CogInfoCom systems. The learner's engagement in language-learning dialogues is strengthened by the robot's non-verbal behaviour such as nodding, gesturing and face-tracking, and the learner's level of interest can be estimated by monitoring the learner's non-verbal signals through gaze-tracking and proximity measurement. In this paper we discuss new approaches in which robotassisted language learning is further enhanced by the use of CogInfoCom channels to extend the learner's engagement and interest. We focus on the robot's use of Wikipedia-based information access to greatly extend the range of topics that the robot can talk about during language-learning spoken dialogues, and describe natural language processing techniques to generate questions from Wikipedia information, so the robot can ask new questions about the learner's preferred topics. We predict that the learner's interest level is increased when the learner can navigate in Wikipedia to dialogue topics of personal interest rather than being restricted to topics chosen by the teacher.