Open Competition 3 - Information Technology

Focused Individualized Realistic Student Tutoring (FIRST)

Sponsor: Carnegie Speech Company

• Project Performance Period: 5/1/2004 - 4/30/2007
• Total project (est.): $2,308,787.00
• Requested ATP funds: $1,909,982.00

A good teacher not only notices when the student makes a mistake but more importantly recognizes the underlying reason for the mistake, and guided by previous experience, explains it in the way that the student is most likely to understand. Carnegie Speech Company plans to build similar insight and flexibility into a tutoring software system aimed at teaching languages and basic mathematics. The project addresses a growing need in two key areas of education: English and math. The number of people wishing to learn English as a second language is currently estimated at about 500 million worldwide and growing, but there is a profound shortage of qualified bilingual education teachers. (In 1997 only 2.5 percent of the teachers instructing classes in English as a Second Language actually had a degree in ESL, according to Carnegie Speech.) Meanwhile some government estimates put the cost of adult illiteracy in the U.S. as high as $300 billion per year in lost productivity. Basic math is in similar straits - the Department of Education estimated that 89 percent of the jobs created between 1999 and 2000 required college-level math and reading skills, but only 50 percent of the students entering the workforce had those skills. But to be really effective, instruction has to be individualized - it is well known among educators that different individuals learn through different mechanisms. Some are more visually oriented, for example, while others need to hear the lesson. Carnegie Speech proposes to develop a general software architecture for computer tutoring systems to support classroom teachers that will be able to pinpoint learner errors on multiple levels; present targeted corrections in real time; and have the flexibility to switch focus when the system, for example, realizes that although the student has made a lexical error in choosing the wrong word, there is a more important underlying grammatical confusion to clear up first. The system will use speech synthesis and speech recognition to interact with learners in a more natural way than typed text. The FIRST (Focused Individualized Realistic Student Tutoring) system also will be able to adapt itself to different learning modes when it recognizes that the student comprehends better with one technique than another. In addition to language skills and basic math, according to the company, the FIRST architecture should be applicable to more specialized job-training tasks for industry. The project must overcome several significant challenges, including dealing with machine errors introduced by the speech recognition system, correctly pinpointing the nature of learner errors, and maintaining a balance between switching focus well enough to correct the most pressing errors, but not so often that the student becomes confused. The company sought ATP support for the work because other sources of funding considered the project to be at too early a stage to accept the risk. If successful, FIRST could help the U.S. make sizable inroads in reducing the annual $300 billion expense of adult illiteracy and provide an important new classroom tool for teachers struggling to meet the requirements of the No Child Left Behind Act of 2001. In addition, in improving the education of the vast number of people learning English worldwide, FIRST can become a significant export and help facilitate international business with the United States.

For project information:

Maxine Eskenazi, (412) 268-3858
max@cs.cmu.edu

ATP Project Manager

Barbara Cuthill, (301) 975-3273
barbara.cuthill@nist.gov