Computerized Speech Tutor

Wiktor Gonet
UMCS, Lublin

ABSTRACT

The growth of speech recognition techniques based on systems trained on large accented speech corpora creates the conditions for the development of goal-oriented software packages that will be applied to solving specific pedagogical goals at various levels of foreign language teaching. In such a system, the learner's speech can be analyzed by the system so that mispronounced phones will be filtered out and explicitely shown to the learner on the computer screen by means of user-friendly graphic interface. The system will provide training loops for specific problems and correction will be greatly facilitated by conspicuous graphic and verbal feedback. First attempts to design such systems have already born fruit in the form of working pilot programs; one such program will be demonstrated.