EUROSPEECH 2001 Scandinavia
The issue of efficient language documentation, particularly with regard to minority and endangered languages, has gained in importance in recent years, as witnessed by several major funding programmes and other human language technology initiatives in the field. An application of finite state technologies to the processing of lexical tone variation in annotated corpora of African languages is described. It is shown that finite state transducers can be constructed which not only provide adequate models for contextual variation in lexical tone (including automatic downstep, downdrift, and tonal assimilations, but also that the transducers provide intuitively satisfying explications of prosodic concepts in `metrical phonology' in terms of oscillations (iterative transitions). The technique has both theoretical value in formalising typological differences in African lexical tone languages and practical value in automatically generating markup enhancements for concordance-based corpus analysis and for fundamental frequency prediction in pitch modelling.
Bibliographic reference. Gibbon, Dafydd (2001): "Finite state prosodic analysis of african corpus resources", In EUROSPEECH-2001, 83-86.