Simulation of Conversation Behavior in Case of Impaired Telephone Transmission

• Applicant: Professor Dr.-Ing. Sebastian Möller
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing ; General and Comparative Linguistics, Experimental Linguistics, Typology, Non-European Languages ; Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
• Term: from 2016 to 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 320253669

With the advent of speech transmission over the internet, impairments such as delay and packet loss affect the conversation behavior of the interlocutors. As an example, end-to-end delay affects turn-taking and packet loss degrades intelligibility of the transmitted messages. Thus, both types of impairments provoke follow-up requests, meta communication and barge-in, and degrade perceived quality by influencing conversation behavior.It is the aim of the present research project to describe the conversation behavior of two interlocutors who communicate via an impaired telephone channel using formal models, and hereby make it accessible to an automatic simulation. Agenda-based models which are successful in human-machine interaction are transferred to human-human dialogues and augmented with descriptions of turn-taking with delayed transmission and misunderstanding handling due to packet loss. The simulation is carried out on the levels of text, intonation, semantics and pragmatics using linguistic material from conversation tests. During the simulation, new conversations are generated based on this material dependent on the transmission channel delay and packet loss. Simulated dialogues are then compared to empirically-observed ones for evaluation purposes. A model for predicting conversation quality is derived from these data. The simulation approach is thus not only useful for research into basic aspects of human communication behavior, but can also be used for quality engineering of communication networks.

• DFG Programme: Research Grants