Data Description and Specification




Please note that Hindi-English and Bengali-English train and test data have been updated on February 16, 2021. Please discard the data if you have downloaded before February 16, 2021. Make sure you download the updated data from the links below in this page.

Please note the followings before downloading the data:




Blind Test Sets

# Subtask Link To Download
1 Subtask 1 Main    Mirror1 (password-free)
2 Subtask 2 Main    Mirror1 (password-free)

Train and Test Sets

# Language Duration (hrs) Sampling Rate Domain Brief Description License Download Link
1 Hindi Train 95.05 8kHz Collection of Stories Click to View Link Main    Mirror1 (password-free)
Test 5.55 8kHz Collection of Stories Click to View Link Main    Mirror1 (password-free)
2 Marathi Train 93.89 8kHz Collection of Stories Click to View Link Main    Mirror1 (password-free)
Test 5 8kHz Collection of Stories Click to View Link Main    Mirror1 (password-free)
3 Odia Train 94.54 8kHz Healthcare, Agriculture, Finance Click to View Link Main    Mirror1 (password-free)
Test 5.49 8kHz Healthcare, Agriculture, Finance Click to View Link Main    Mirror1 (password-free)
4 Tamil Train 40 16kHz General Click to View Link Click to Download (ta-in-Train)
Test 5 16kHz General Click to View Link Click to Download (ta-in-Test)
5 Telugu Train 40 16kHz General Click to View Link Click to Download (te-in-Train)
Test 5 16kHz General Click to View Link Click to Download (te-in-Test)
6 Gujarati Train 40 16kHz General Click to View Link Click to Download (gu-in-Train)
Test 5 16kHz General Click to View Link Click to Download (gu-in-Test)
7 Hindi-English Train 89.86 16kHz Tech Lecture in computer science Click to View CC BY-SA 4.0 Main    Mirror1 (password-free)
Test 5.18 16kHz Tech Lecture in computer science Click to View CC BY-SA 4.0 Main    Mirror1 (password-free)
8 Bengali-English Train 46.11 16kHz Tech Lecture in computer science Click to View CC BY-SA 4.0 Main    Mirror1 (password-free)
Test 7.02 16kHz Tech Lecture in computer science Click to View CC BY-SA 4.0 Main    Mirror1 (password-free)