Multi-episodic perceived quality for one session of consecutive usage episodes with a speech telephony service

Dennis Guse; Benjamin Weiss; Frank Haase; Anna Wunderlich; Sebastian Möller

In: Quality and User Experience, Vol. 2.8, Pages 1-16, Springer, DE-Berlin, 2017.


The impact of varying performance and its effect on the perceived quality is an important aspect of quality of experience. Especially for service providers, it is important to understand how the perceived quality of a user, who is interacting with their services repeatedly, evolves. Repeated-use of a service is actually common for telecommunication services, such as speech telephony. For telecommunication services, it is very likely that a user encounters varying performance, for example due to varying network load conditions. For repeated-use taking place in a usage period covering several days, initial work on the formation process of the so-called multi-episodic perceived quality (i.e., distinct, meaningful interactions, denoted as usage episodes) has been conducted. In this paper, we present our work investigating the multi-episodic perceived quality for usage episodes taking place in one session (i.e., in a time frame of up to 45 min) with a speech telephony service. We investigate two aspects and their impact on the multi-episodic perceived quality: (a) increasing the number of low performance (LP) usage episodes and (b) varying the temporal position of LP usage episodes towards the assessment of multi-episodic perceived quality. Additionally, the impact of the user’s behavioral freedom (i.e., two-party conversation vs. 3rd-party listening) on the two investigated aspects is examined. With regard to (a), the results show that an increasing number of LP usage episodes lead to a reduced multi-episodic perceived quality until saturation is reached. No impact of the user’s behavioral freedom was found. For (b), an impact of the varied behavioral freedom was observed. Varying the temporal position had an impact on the multi-episodic perceived quality for two-party conversation. For 3rd-party-listening, this could only be observed in one out of two cases.

Weitere Links

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence