User Modeling in Spoken Dialogue Systems to Generate Flexible Guidance

Abstract

We address the issue of appropriate user modeling to generate cooperative responses to users in spoken dialogue systems. Unlike previous studies that have focused on a user's knowledge, we propose more generalized modeling. We specifically set up three dimensions for user models: the skill level in use of the system, the knowledge level about the target domain, and the degree of urgency. Moreover, the models are automatically derived by decision tree learning using actual dialogue data collected by the system. We obtained reasonable accuracy in classification for all dimensions. Dialogue strategies based on user modeling were implemented on the Kyoto City Bus Information System that was developed at our laboratory. Experimental evaluations revealed that the cooperative responses adapted to each subject type served as good guides for novices without increasing the duration dialogue lasted for skilled users.

Adaptive Interfaces and Agents

“…Since it is not always this easy to classify the user, and since asking the user for a self-assessment can be awkward and time-consuming, researchers have looked for ways of adapting to the user on the basis of her behavior during the current dialog (see, e.g., Litman & Pan, 2002). A recent example is given in Figure 22.12, which shows a translation of an example dialog conducted with the KYOTO CITY BUS INFORMATION SYSTEM (Komatani, Ueno, Kawahara, & Okuno, 2005). The system adjusts its assessments of three properties of the caller on the basis of each utterance of the caller: level of skill at conducting dialogs with this system; level of knowledge concerning the domain (i.e., Kyoto and its buses); and desire to complete the dialog quickly ("urgency").…”

Achieving rapport with turn-by-turn, user-responsive emotional coloring

“…One illustration is that of monitoring for user frustration and when detected, apologizing or performing other mitigating actions (Klein et al, 2002). A second illustration is that of monitoring for urgency in the user's voice, words, and turn-taking behavior, and when this is detected skipping the confirmation step to give the desired information immediately (Komatani et al, 2005). A third illustration is in the domain of tutoring dialogs, where Forbes-Riley and Litman (2011) showed that monitoring the user's level of uncertainty, which can be inferred from prosody (Pon-Barry, 2008), and responding to uncertain answers with more explanation was better liked and improved learning.…”

Evaluation and usability of multimodal spoken language dialogue systems

“…Bernsen (2003) describes an application of individual on-line user modelling to the hotel reservation task in an in-car SLDS. Komatani et al (2003) describe an application of generic on-line user modelling which adapts the system's information level to user experience with a bus information system. General on-line user modelling is an active research area.…”