| Date: | Thursday, March 2 |
| Time: | 11:00am-12:00pm |
| Location: | CS Conference Room (Mudd) |
| Speaker: | Linda Bell |
| A senior speech technology researcher at the R&D division
at TeliaSonera. Linda's interested in how users adapt their linguistic
behavior when they interact with spoken and multimodal dialogue systems,
questions addressed in several projects at KTH and Telia. Linda has been
working in the field of multimodal dialogue systems since 1998, and
received her PhD in speech communication from KTH in 2003. |
|
| Speaker: | Joakim Gustafson |
| A senior speech technology researcher at the R&D
division at TeliaSonera. Starting with the Waxholm system at KTH, Joakim
has more than ten years of experience in the design and development of
spoken and multimodal dialogue systems. Since joining TeliaSonera, he
has managed several speech projects, among them the EU-IST-financed
NICE. Joakim received his PhD in speech communication from KTH in 2002. |
|
| Abstract: |
In this talk, we will describe some of our experiences over the last 8
years in developing four different Swedish multimodal dialogue systems.
The August system, an information kiosk, was exposed to the general
public and put on display in an exhibition area downtown Stockholm. The
AdApt system, a virtual broker, helped users access real estate
information from a database of Stockholm apartments. The Pixie system, a
virtual home helper, was available as part of a museum exhibition at the
Technical museum of Stockholm for almost two years. Finally, the NICE
system, a speech-enabled computer game, was made available to school
children of different ages.
Our data collections and subsequent analyses have focused on how users adapt their language and dialogue behavior when they interact with the multimodal dialogue systems. Since several of our systems were used by children and adolescents, we have studied the differences between adult and children behavior in human-computer interaction. Moreover, we have made efforts to analyze how the system design influenced the users' behavior at different levels. |