Consistent and Grounded Dialogue Response Generation. Generating responses to dialogues such that the responses are logically and factually consistent with the previous history of the dialogue and supplied external context

Conversational experiences are becoming more prevalent in software applications – from Alexa and Google assistant to Siri and Cortana – though the quality degrades when the automated dialogue agent must refer to or recall specific information. These systems are often forgetful, nonsensical, contradictory, repetitive, or hallucinatory, which impacts the user experience. This project aims to improve the consistency of responses generated by dialogue agents, thereby improving user engagement and trust in the software bot. Microsoft has many conversational experiences embedded in their software and solutions which can benefit from such a solution, like Cortana, Customer Care, Bot Framework, and Azure Cognitive Services.

Faculty Supervisor:

Frank Rudzicz

Student:

Partner:

Microsoft Canada

Discipline:

Computer science

Sector:

Information and cultural industries

University:

University of Toronto

Program: