tags: - k2 - icefall - audio - automatic-speech-recognition language: en datasets: - speechcolab/gigaspeech