Training and Evaluating an E-tutor

Abstract

This work generates an ontology for Operating Systems which map the students′ lexicon against ours. Their exploration of concepts generates data to train a system that focuses or broadens the interaction with the student to form a conversation. The data we are collecting will also be used to verify the knowledge base design itself by using inductive logic and statistical techniques to find patterns of misunderstanding. The tools allow students to create their own models as well as to ask questions to a Chatbot built on a teacher′s model. The students can mark their work against this teacher′s model. We start from a pedagogical approach and then evaluate use of our Chatbot to look for causal patterns in the learning material that can be used in future for automatic trouble shooting via a computer initiated dialogue that can discuss the structure of a subject with a student based on that particular student′s level of understanding.