April 1, 2003

Students Develop Innovative Prototype Speech Applications, Part II

By Dr. James A. Larson

A year ago, I reported on the cool applications developed by some university students in the May/June issue of Speech Technology Magazine, [http://www.speechtechmag.com/issues/7_3/cover/748-1.html ]. Not to be outdone, students this year developed a whole new crop of speech applications.

Students at Portland State University created an “author once and use in many environments” system. A user interface is authored by specifying a state transition system using a special XML language called GraphXML. Three transducers convert the specification to:

The students resolved issues such as:

Ioan Bec, Gabrielle Gilbertsen, Mojdeh Givi, Jeff Hodel and Sean McMillen, demonstrated that an “author one and use in many environments” is possible, but agreed that developers should modify the generated user interfaces to improve usability preference and performance scores.

Students at the Oregon Graduate Institute, Oregon Health and Science University created several VoiceXML speech applications as the end result of performing seven user-centric activities:

1. Ethnography — Record subject activities with a view to identify new applications that make life easier for the subject.

2. Survey — Construct and conduct a survey of subjects to evaluate usefulness of several candidate applications.

3. Focus group — Present 3-5 subjects with three simulated application scenarios and solicit their feedback about the usefulness of each application.

4. Conceptual model — Work with subject to construct a conceptual model of the application that contains the names of application objects and verbs a user might speak to manipulate the object.

5. Dialog scenario — Work with the subject to create a state transition diagram representing a simple, system-directed dialog of the use of the application.

6. Wizard of Oz test — Subject telephones student who pretends to be the application by asking the subject questions in a dialog scenario.

7. Usability test — Subject interacts with an implemented VoiceXML speech application with measurements of subject performance and preference recorded.

Students created prototype voice user interfaces to the following applications:

Child Activity Reporter by Rajshree Sankaran and Kuan-Hsing Chen — Parents call to review their child’s activities at daycare.

Hospital information and medical advice system by Aaron Bentley, Yi Huang, and Madhavi Pullela — information on hospitals based on their location, hours of operation, automated medical advice and transfer to a live advice nurse.

Kid’s Activity Line by Kim Paulson, Yiyang Fei and Tawanna Dillahunt — child selects stories to listen to, verbal games to play and educational activities for learning skills.

Grocery list organizer by Bharat and Parulata Trivedi — enter and access grocery items for later purchase at a supermarket.

Task reminder by Bob Myers and Manisudha Yalamanchi — verbal to-do list retrieved and updated via a telephone.

To Do List by Rebecca Lundsford and Helen Ross — verbal to-do list retrieved and updated via a telephone.

Bank Automation Application by Kiran Muthabatulla and Jolly Mehta — pay bills via the telephone.