ReTiCo: An open-source framework for modeling real-time conversations in spoken dialogue systems

Thilo Michael, Sebastian Möller

In: Tagungsband der 30. Konferenz Elektronische Sprachsignalverarbeitung 2019. Elektronische Sprachsignalverarbeitung (ESSV-2019) March 6-8 Dresden Germany Pages 134-140 ESSV ISBN 978-3-95908-157-3 TUDpress, Dresden 2019.


In this paper we present ReTiCo, a python-based programming frame-work that utilizes the concepts of incremental processing to create interactive spo-ken dialogue systems and simulations of conversational behavior. In contrast toalready existing toolkits like InproTK, our framework allows for quick visual cre-ation of complex networks and is able to save and load incremental networks forsimulations, automated testing as well as analysis. It is focused on simplifying theuse of incremental modules to make research on specific tasks of an incrementaldialogue system (like real-time speech signal processing) and on the interactionbetween different incremental modules easier.We make this framework accessible as open source so that it can be used in researchon spoken dialogue systems and conversation simulation.

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz