Collecting and evaluating speech recognition corpora for 11 South African languages

Abstract

The authors describe the Lwazi corpus for automatic speech recognition (ASR), a new telephone speech corpus which contains data from the eleven official languages of South Africa. Because of practical constraints, the amount of speech per language is relatively small compared to major corpora in world languages, and they report on their investigation of the stability of the ASR models derived from the corpus. They also report on phoneme distance measures across languages, and describe initial phone recognisers that were developed using this data. They find that a surprisingly small number of speakers (fewer than 50) and around 10 to 20 hours of speech per language are sufficient for the purposes of acceptable phone-based recognition.