To support Indian language voice technologies, the Bill & Melinda Gates Foundation and the German Development Cooperation have provided financial support through their respective programs. The next generation of internet users may be able to bridge the digital divide with the help of voice technology in their native tongues. The project aims to create and make available datasets for the development of voice technology in nine Indian languages: Bhojpuri, Maithili, Maghadi, Hindi, Bengali, Kannada, Telugu, and Marathi. FAIR Forward has started a project in India to use voice technology to its full potential. More than 11,000 hours of high-quality, gender-balanced voice datasets will be collected as part of the project. The data will be available to Indian academics, start-ups, researchers, and developers free of charge.
Key Highlight:
- The Bill and Melinda Gates Foundation and the German Development Cooperation initiative (FAIR Forward – Artificial Intelligence for All) have awarded USD 2 million in grants to support open-source voice technologies in Indian languages.
- The project aims to create and make available datasets for the development of voice technology in nine Indian languages: Bhojpuri, Maithili, Maghadi, Hindi, Bengali, Kannada, Telugu, and Marathi.
- More than 11,000 hours of high-quality, gender-balanced voice datasets will be collected as part of the project.
- The data will be available to Indian academics, start-ups, researchers, and developers free of charge.
The Bill and Melinda Gates Foundation and the German Development Cooperation initiative (FAIR Forward – Artificial Intelligence for All) have awarded USD 2 million in grants to support open-source voice technologies in Indian languages to a team led by Prasanta Kumar Ghosh, Associate Professor at the Department of Electrical Engineering at the Indian Institute of Science (IISc).
Human-machine interaction has become increasingly important in the age of mobile phones and other connected devices because it allows users to gain access to information, services, and resources. However, many low-income people cannot use digital technologies effectively because of literacy, skills, poverty, gender, and other socio-economic biases. People in low-income areas are still excluded from the benefits of this technological revolution, even though voice technologies have made significant progress with the introduction of digital assistants like Alexa, Cortana, Siri, Google Assistant, and others. As a result of existing gender gaps in access, education, rights, and empowerment, the digital divide widens even further for women, especially those with low income. New artificial intelligence and machine learning models largely ignore these excluded groups’ languages, dialects, and accents. For the next generation of internet users, voice technology in local languages targeted at these underserved groups can help close the digital divide.
The researchers at IISc have set out to create and make available datasets to develop voice technologies in nine Indian languages: Bhojpuri, Maithili, Maghadi, Hindi, Chhattisgarhi, Bengali, Kannada, Telugu, and Marathi, with this objective in mind. There is a shortage of publicly available training data for developing such voice technologies in Indian languages and local innovation. In addition, they concentrate on the dialects and languages of the most lucrative and developed economies, favoring urban and well-educated users. For less literate and marginalized populations, collecting open voice data will help local AI ecosystems and give millions of people access to services they haven’t had before, whether in agriculture, education, health care, or other fields.
An open-source AI-based solution will be created using freely available voice data sets that can be used to train machine learning algorithms. By contributing to NLTM, the project will help unlock the vast potential of voice technology, which has been largely untapped. NLTM is an initiative of India’s Ministry of Electronics and Information Technology.
By collecting more than 11,000 hours of high-quality voice datasets from both men and women in nine Indian languages, the project aims to help poor farmers and women in the finance and agricultural industries. To develop text-to-speech applications in the same nine Indian languages, the FAIR Forward investment will support the collection of nearly 1,000 hours of gender-balanced, high-quality speech recordings from voice artists. To encourage innovation and academic activity in the development of regional voice technologies in India, the datasets will be made openly and freely available to Indian academics, start-ups, researchers, and developers. As a result, India will benefit from a more robust technology ecosystem as the creation and use of voice technologies become more accessible to a broader range of people.
https://respin.iisc.ac.in
https://syspin.iisc.ac.in