Publication
MAT: a tool for L2 pronunciation errors annotation
Renlong Ai; Marcela Charfuelan Oliva
In: Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14). International Conference on Language Resources and Evaluation (LREC-2014), May 26-31, Reykjavik, Iceland, ISBN 978-2-9517408-8-4, European Language Resources Association, 2014.
Abstract
In the area of Computer Assisted Language Learning(CALL), error-annotated second language(L2) learner data has been an important type of resource for training automatic error detection and also for testing and evaluating. However, the acquisition of such data is difficult due to the annotation work, which has to be manually done by linguists or phoneticians. This paper describes MAT (MARY Annotation Tool), a platform-independent tool for L2 pronunciation errors annotation. It aims at providing an easy and fast annotation process via a comprehensive and friendly user interface. The tool is web-based and covers most of the common problems in pronunciation training. Errors are categorized in phoneme, syllable, word and sentence levels and can be configured upon demand. The tool is based on the MARY TTS, from which it uses the components: text analyser (tokeniser, syllabifier, phonemiser), speech signal processor and phonetic aligner. Annotation results are stored in XML format, which is easy to process and analyze, and also possible to transform to other formats.