Acquiring Reusable Multilingual Phonotactic Resources

Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC 2004)

Abstract

This paper presents a fully automatic procedure for acquiring reusable phonotactic resources from syllable annotated data. The procedure makes use of a regular inference algorithm and the acquired resources are stored in a specialised XML representation. The technique is then extended to support acquisition from phoneme labelled data while providing a semi-automatic annotation system assisting user annotations of phoneme labelled data with syllable boundaries.