Learner Answer Assessment in Intelligent Tutoring Systems

Truly effective dialog and pedagogy in Intelligent Tutoring Systems will only be achievable when systems are able to understand the detailed relationships between a learner’s answer and the desired conceptual understanding. This thesis describes a new paradigm and framework for recognizing whether a learner’s response to an automated tutor’s question entails that they understand the concepts being taught. I illustrate the need for a finer-grained analysis of answers than is supported by current tutoring systems and describe a new representation for reference answers that addresses these issues, breaking them into detailed facets and annotating their relationships to the learner’s answer more precisely. Human annotation at this detailed level still results in substantial inter-annotator agreement, 86.1%, with a Kappa statistic of 0.728. I present current efforts to automatically assess learner answers within this new framework, which involves training machine learning classifiers on features extracted from dependency parses of the reference answer and the learner’s response and features derived from domain-independent lexical statistics. The system’s performance, 75.5 % accuracy within domain and 65.9% out of domain, is very encouraging and confirms the approach is feasible.