Design and Implementation of Conversational Agents for Harvesting Feedback in eLearning Systems

Abstract

Traditionally conversational interfaces, such as chatbots, have been created in two distinct ways. Either by using natural language parsing methods or by creating conversational trees that utilise the natural Zipf curve distribution of conversations using a tool like AIML. This work describes a hybrid method where conversational trees are developed for specific types of conversations, and then through the use of a bespoke scripting language, called OwlLang, domain knowledge is extracted from semantic web ontologies. New knowledge obtained through the conversations can also be stored in the ontologies allowing an evolving knowledge base. The paper describes two case studies where this method has been used to evaluate TEL by surveying users, firstly about the experience of using a learning management system and secondly about students’ experiences of an intelligent tutor system within the I-TUTOR project.