A Conversational Agent to Shift Students' Affect State

摘要

When a student is feeling negatively while doing some required tasks in a learning environment, in this case, a reading assessment exercise, a conversation with the tutor agent can be initiated to engage the student in an affective text-based dialogue as a means of intervention. Such dialogue can revolve around everyday, commonsense topics that may be related to the reading material or exercise at hand. A semantic ontology populated with commonsense concepts from existing knowledge sources, specifically Con-ceptNet and SenticNet, provides the conversational agent with the candidate set of topics for discourse. To facilitate the validation of the dialogue system with children, the conversational tutor agent was integrated into a learning environment platform that supports reading of short stories.