Skip to main content

Technology and Architecture for Spoken Dialog Systems

Final Activity Report Summary - CASA (Technology and Architecture for Spoken Dialog Systems)

Conversational systems allow individuals to interact with computer systems using spoken natural language in order to perform specific tasks as they would with human agents. Examples of interactions with a conversational system include complex tasks like asking specific questions about sports, weather, news, stock quotes, executing bank transactions, planning travels, routing the user's call to a human operator.

Research in this field is hampered by the complexity of such systems requiring multidisciplinary expertise. On the other hand, recent trends in speech-enabled services have seen an increase in exploitation of such technology.

We have created a spoken dialog system technology platform based on off-the-shelf technology and standard based VXML platform. The telephony architecture supports both human-machine (HM), human-human (HH) and three-way call (e.g. Wizard-of-Oz experiments) both using PSTN and soft IP phones.

We have developed tools for the logging, monitoring, data mining of telephone conversations, both HM and HH. The mining of dialog conversations provide standalone statistics of speech events such as confidence score statistics over dialogs and semantic forms. The CASA tools provide an XPATH search functionality that allows to precompile evaluation queries or let the user design its own query.

We have prototyped several spoken dialog systems with in-lab and real users and evaluated the spoken dialog systems and ascertained the user acceptance of the conversational systems from the user point-of-view. Currently, the spoken dialog platform we have developed is supporting the education and research of many people in the lab and as well as the collaboration with external research groups.

The goal of the CASA project has been to establish a state-of-the-art spoken language dialog lab to foster both education and research and bridge the gaps between education, research and technology.