KALAKA-3: a database for the recognition of spoken European languages on YouTube audios

摘要

This paper describes the main features of KALAKA-3, a speech database specifically designed for the development and evaluation of language recognition systems. The database provides TV broadcast speech for training, and audio data extracted from YouTube videos for tuning and testing. The database was created to support the Albayzin 2012 Language Recognition Evaluation, which featured two language recognition tasks, both dealing with European languages. The first one involved six target languages (Basque, Catalan, English, Galician, Portuguese and Spanish) for which there was plenty of training data, whereas the second one involved four target languages (French, German, Greek and Italian) for which no training data was provided. Two separate sets of YouTube audio files were provided to test the performance of language recognition systems on both tasks. To allow open-set tests, these datasets included speech in 11 additional (Out-Of-Set) European languages. The paper also presents a summary of the results attained in the evaluation, along with the performance of state-of-the-art systems on the four evaluation tracks defined on the database, which demonstrates the extreme difficulty of some of them. As far as we know, this is the first database specifically designed to benchmark spoken language recognition technology on YouTube audios.