Second European Conference on Speech Communication and Technology

Genova, Italy
September 24-26, 1991


Macro and Micro Features for Automated Pronunciation Improvement in the Spell System

Edmund Rooney (1), Steven Hiller (1), John Laver (1), Maria-Gabriella Di Benedetto (2)

(1) CSTR, University of Edinburgh, UK
(2) INFOCOM Dept. , University of Rome, Italy

The analysis of macro (prosodic) and micro (segmental) features is described for a workstation designed to improve the pronunciation of English, French and Italian by non-native speakers. The SPELL workstation is intended to be a teaching device aimed at intermediate ability foreign language learners. Audio and visual aids will be used to help students improve their general intelligibility within a basic teaching paradigm called DELTA (Demonstrate, Evaluate Listening, Teach and Assess). Prosodic analysis will apply to the features of intonation, stress and rhythm. A phonological approach is used for intonation which provides a well-structured system of contrasting units that correlate with discrete linguistic functions. A more limited approach to the prosodic phonology of stress and rhythm will be taught in the SPELL system by manipulating vowel quality and segmental duration. The micro feature analysis will focus on vowel contrasts, using a distinctive feature approach to characterize non-native vowel pronunciation. Acoustic properties are sought which will be speaker-independent. Keywords: Computer-aided Language Learning; Pronunciation, Prosody, Articulatory Phonetics, ESPRIT

Full Paper

Bibliographic reference.  Rooney, Edmund / Hiller, Steven / Laver, John / Benedetto, Maria-Gabriella Di (1991): "Macro and micro features for automated pronunciation improvement in the spell system", In EUROSPEECH-1991, 985-988.