The Austrian Language Project (T-ALP) aims to improve automatic speech recognition of Austrian German. Specifically, it focuses on Austrian German as it is typically spoken in the media such as TV and radio as well as podcasts but, at least at present, does not try to cover the wide dialectal and sociolectal variety of German spoken in Austria. The project was presented at the conference Building a European Cultural Backbone (ECB2022) at the Stadtwerkstadt in Linz, Austria.

I initially became involved in this project when I met Alexander Baratsits, one of the founders of the Cultural Broadcasting Archive (CBA). The …

Note: This is about a recipe for the original, English-only Common Voice corpus released in 2018. Subsequent releases with data for various languages are not compatible with this recipe.

Recently, Mozilla published a first version of their Common Voice corpus. It consists of speech prompts from an unknown number of speakers (no speaker IDs are provided due to privacy concerns, see this forum thread). About 254 hours have been validated by multiple listeners. The data has been split into pre-defined training, development and test sets.

I created an initial Kaldi recipe for the corpus as a research exercise. It has …