Title

Three New Corpora at the Bavarian Archive for Speech Signals - and a First Step Towards Distributed Web-Based Recording

Authors

Christoph Draxler (BAS Bavarian Archive for Speech Signals, University of Munich, Germany) 

Florian Schiel (BAS Bavarian Archive for Speech Signals, University of Munich, Germany)

Session

SO1: Large Projects-Initiatives For Speech Corpora

Abstract

The Bavarian Archive for Speech Signals has released three new speech corpora for both industrial and academic use: a) Hempels Sofa contains recordings of up to 60 seconds of non-scripted telephone speech, b) ZipTel is a corpus with telephone speech covering postal addresses and telephone numbers from a real world application, and c) RVG-J, an extension of the original Regional Variants of German corpus with juvenile speakers. All three corpora were transcribed orthographically according to the SpeechDat annotation guidelines using the WWWTranscribe annotation software. Recently, BAS has begun to investigate performing large-scale audio recordings via the web, and RVG-J has become the testbed for this type of recording.

Keywords

Speech database, Corpus, Telephone speech, Real-World application

Full Paper

20.pdf