Students Develop Innovative Prototype Speech Applications
Student teams are very creative when asked to design and implement speech applications of their choice. Here are some of the prototype speech applications recently implemented by students at Georgia Institute of Technology, Washington State University, Portland State University, and Oregon Health and Sciences University.
Talking Recipe Book. The talking recipe book enables cooks to access recipe instructions by voice while their eyes and hands are busy preparing a meal. The talking recipe book informs cooks which ingredients are needed and provides step-by-step instructions to prepare the dish, and, if necessary, how long and at what temperature to cook the dish. Two teams of students, one from Georgia Tech and another from Washington State were given 15 minutes to convert a recipe to the RecipeXML languages they designed. Ingredients were provided for each team, who interacted with their talking recipe books to guide them in preparing the dish. Both teams ate their experiments. My (formerly secret) recipe for bean dip tasted pretty good.
Sports Score. Callers ask for recent scores and statistics about NFL teams. A Portland State team dynamically scraped data from an existing sports Web page to obtain scores and statistics about NFL teams. Callers ask for the latest scores and statistics of their favorite teams
Word Games. A Portland State team developed several word games, including "Name That Tune" and "Jeopardy." These word games can be used to train novice users in how to use voice applications and provide a fun entertainment session playing the games. The word games could also be used to train a speech recognition engine, replacing the boring training sessions that many users dislike with a more enjoyable experience.
Used Car Prices. A Portland State team used a VoiceXML form to solicit description of a used car. The system scraped information from a visual Web site, and reported the estimated price to the caller. Callers use the system to get realistic estimates for used cars.
Please Advise. This application automatically generates VoiceXML menus to diagnose a problem and make appropriate recommendations. A Portland State team used graphs to represent questions (nodes) and answers (arcs), by structuring the graph so that the answer to one question leads to a more specific question. The graph represents a sequence of questions that embody a dialog in which the user self-diagnoses a problem and arrives at a possible solution. Example graphs might contain questions and answers leading to:
- Diagnose the cause of an equipment failure
- Provide a tentative diagnosis of a common medical condition
- Capture information about an investment goal and recommend an investment strategy
- Advise the caller of missing resources needed by a computer in order to run a new application
The graph structure is converted to a sequence of VoiceXML menus that enable callers to answer the questions by phone and receive helpful advice. Voice applications for seniors. During a ten-week course at the Oregon Health and Sciences University, students used a variety of ethnographic and interviewing techniques to identify a collection of useful speech applications. Students involved more senior citizens in the design and evaluation of several VoiceXML applications, including:
- Event information for a senior citizen center
- Reminder to take medicine at appropriate times
- Furniture comparison-shopping
- Book availability
- Automobile part comparison-shopping
- Comparison of insurance policy rates
Send brief descriptions of your innovative student projects. I'll post them on my Web site so students can review them and generate even more creative projects. I can hardly wait to see…make that listen to…the projects created by other students.
James A. Larson is an adjunct professor at Portland State University and Oregon Health Sciences University. He can be reached at email@example.com and his Web site is www.larson-tech.com