HINDI asr challenge

Speech Lab, IIT Madras, presents the Automatic Speech Recognition (ASR) in Hindi challenge. This challenge is a part of the National Language Translation Mission funded by MeitY. It aims towards helping and encouraging the advancement of ASR in Indian Languages. We plan to have a series of challenges with increasing difficulty in different Indian languages, and release appropriate data with each challenge. In the first few challenges, we will release everything including source codes etc, so that start-ups/Universities/Research-Labs without previous experience in ASR can also participate and get familiar with it.

Challenge overview

Recent advancements in Speech technology have shown that ASR systems can work on par with humans. To build an efficient ASR system, it would require large amounts of training data and high-end computational resources.

However, when it comes to Indian languages, not everyone, especially academic institutions and startups, have access to these resources. As a part of this challenge, we will be releasing speech data in Hindi. Everyone who participates in this challenge will then be free to use this data for research purposes.

Data Set and Baseline recipes

The data set comprises of Hindi read speech data along with the corresponding transcriptions. The text data was crawled from newspapers, and then volunteers were asked to read them. It covers genres like politics, sports, entertainment, etc. The following data sets will be released for this challenge:

• Train set - 40 hours

• Development set - 5 hours

• Evaluation set - 5 hours

Lexicon, baseline models, results and recipes to replicate the baseline experiments will also be made available.

Closed Hindi-ASR Challenge: Only the training data distributed as part of the challenge can be used to train the models (both acoustic and language models). Please do not use dev set data.

Open Hindi-ASR Challenge: You can use any external/additional data to train the acoustic and language models.

How to Participate

• Enroll yourself by registering on this link: Register Now!

• Registering on the above link provides access to the user license and download the training and test data for Hindi challenge

Submit results: Use submission portal submit your results.

• The submission portal will open on 21st of August 4th of September and closes at midnight on 30th of August 13th of September (midnight anywhere in the world, i.e., 12pm UTC on August 30 September 13)

• Submissions should include the ASR output produced by the system and a brief description of the system (see the documentation for further instructions about formatting and the submission procedure)

• Participating teams can submit a maximum of 10 submissions per team

• Results will be displayed on a leader board throughout the period that the submission site is open

Baseline results and recipes

• Baseline results and scripts can be found here

• You can check out our baseline models on Binder here.

• Check out this link to work with Docker

Important Dates

• Release of training data (40 hours), development data (5 hours), lexicon and baseline system: July 6, 2020

• Evaluation data released and opening of submission site: August 21, 2020 September 4, 2020

• Closing of submission site: August 30, 2020 September 13, 2020 (midnight anywhere in the world, i.e., 12pm UTC on August 30, 2020 September 13, 2020)

• Announcement of results: September 4, 2020 September 18, 2020