Developing corpus interoperability for phonetic investigation of learner corpora

Abstract : Although automatic analysis and computer-aided annotation tools are being developed, spoken learner corpora are still smaller and less numerous than written learner corpora. This chapter gives a critical overview of some of the phonetic research questions addressed by spoken learner corpora in relation to their annotation schemes and software. Some of their annotation schemes and guidelines are presented and assessed. Corpus design and tools are discussed in relation to some two of the challenges of spoken learner corpora: comparability of data and the potential contribution to prosodic modeling. It is argued that reusability of annotated spoken data and critical statistics should be the real order of the day.