Technology Enhanced Learning Using Humanoid Robots

Abstract

In this paper we present a mixture of technologies tailored for e-learning related to the Deep Learning, Sentiment Analysis, and Semantic Web domains, which we have employed to show four different use cases that we have validated in the field of Human-Robot Interaction. The approach has been designed using Zora, a humanoid robot that can be easily extended with new software behaviors. The goal is to make the robot able to engage users through natural language for different tasks. Using our software the robot can (i) talk to the user and understand their sentiments through a dedicated Semantic Sentiment Analysis engine; (ii) answer to open-dialog natural language utterances by means of a Generative Conversational Agent; (iii) perform action commands leveraging a defined Robot Action ontology and open-dialog natural language utterances; and (iv) detect which objects the user is handing by using convolutional neural networks trained on a huge collection of annotated objects. Each module can be extended with more data and information and the overall architectural design is general, flexible, and scalable and can be expanded with other components, thus enriching the interaction with the human. Different applications within the e-learning domains are foreseen: The robot can either be a trainer and autonomously perform physical actions (e.g., in rehabilitation centers) or it can interact with the users (performing simple tests or even identifying emotions) according to the program developed by the teachers.