Skip to main content Skip to main navigation


Automatic Context-Relevant Off-Activity Talk Suggestion for Dialogue Contribution

Omid Moradiannasab
Mastersthesis, Saarland University, Groningen University, 10/2015.


This thesis investigates the possibility of utilizing online resources for off-activity dialogue contribution. It is a first attempt to propose a tool which automatically suggests off-activity talk s in form of some sentences relevant to the dialogue context. We propose four approaches and comparatively evaluate them over two test-sets of open domain and health-related queries in a conversational quiz-like setting. The evaluation results show satisfying performance for some of the proposed approaches. The results suggest that the modular architecture implemented throughout this work provides an applicable and effective system to process dialogue context and suggest relevant off-activity talks.