Leaderboard icon indicating copy to clipboard operation
Leaderboard copied to clipboard

SpeechIO Leaderboard: a large, robust, comprehensive, benchmarking platform for Automatic Speech Recognition.

Results 9 Leaderboard issues
Sort by recently updated
recently updated
newest added

Thanks for the great the work! I'm wondering whether the testing report about academic dataset (especially for the cloud model) is provided? E.g. Microsoft API CER in aishell1 and aishell2

我在[gigaspeech](https://github.com/SpeechColab/GigaSpeech) 上看到[k2_gigaspeech](https://huggingface.co/wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2)模型的效果是10.40 / 10.51 ,如图所示: 而且我看k2 gigaspeech的[测试结果](https://github.com/k2-fsa/icefall/tree/master/egs/gigaspeech/ASR) 也是这个,如图所示: 但是为什么在[Leaderboard](https://github.com/SpeechColab/Leaderboard/blob/master/misc/SpeechColab_ASR_EN_2022_10.png)上的结果对不上呢?如图所示: ![image](https://github.com/SpeechColab/GigaSpeech/assets/22560013/df7369b4-4c1f-4788-b28c-13ba62585d6b)

如果我想上榜你们的排名榜,应该和谁联系

想问一下,我看model目录下面一些demo用到的接口是短语音识别(1分钟以内),如果我的测试语音超过了1分钟,是不是就要用录音文件转写的接口了

比如腾讯云的ASR大模型,头条的豆包ASR大模型等。

按照readme的方式,下载测试集,但是一直卡住了,无法下载到数据,麻烦帮忙看下怎么解决 ![image](https://github.com/user-attachments/assets/c3899bd8-2955-4de9-9f21-a813020f0017)

Hello, I sent an access mail to access the test datasets access key last week, but haven't received a reply yet.

Whether there are plans to push the dataset to huggingface to open source it.