Leave Info.

Release of Baseline Systems for the 5th Oriental Language Recognition Challenge


The OLR challenge series organized by Speechocean and Tsinghua University aim at boosting language recognition technology for oriental languages. Following the success of the past four OLR challenges, the 2020 challenge co-organized by Xiamen University has been opened for registration.


In this challenge, 188 hours of speech recognition corpus covering 18 languages are totally free for each participant. In addition to the conventional "Cross-channel LID" task, “Dialect Identification” and “Noisy LID" are added for the first time.


Specifically, this challenge provides two kinds of baseline systems with Kaldi and Pytorch. The Kaldi baseline includes the version of i-vector and x-vector. And the Pytorch baseline focuses on the x-vector model, which is trained by ASV-Subtools, a new open-source toolkit developed by XMUSPEECH team of Xiamen University. ASV-Subtools can also be used for the task of automatic speaker verification (ASV). All the based systems use Kaldi to conduct the feature extraction and back-end processing.


The Kaldi and Pytorch recipes of these baselines can be downloaded from this website:


And ASV-Subtools can be downloaded through this URL:


Come to participate in this challenge right now:


Leave Information