Potential of Open Voice Technology for India

Providing people with information in their mother tongue is a key driver of economic empowerment and political participation. In India this can become quite a challenge because of the richness and diversity of languages that are spoken. Voice technologies offer novel and efficient ways to tackle this challenge and make society more inclusive. Using the power of Artificial Intelligence (AI), voice technologies have great potential to enable millions of people to access services they are not able to use yet – be it in agriculture, education, health or others. 

Vibrant Indian Voice Technology Community

An active community of researcher, innovators and start-ups across India have tackled this problem. They develop state-of-the-art solutions to in various fields to develop sophisticated systems for automatic speech recognition, speech synthesis or machine translation. This has created early success stories where voice technologies are deployed to improve customer support, search applications and chatbots in vernacular languages. 

Open Dataset in Indian Languages are Key to Technology Development

Too often though developers of voice technology hit a roadblock to fully realize the vision of voice technologies in Indian languages: The lack of free and open voice datasets in Indian languages. Open voice data is the foundation for local AI innovators and researchers to build applications that serve economic and social development. To release the potential of voice-based applications in Indian languages, there is an urgent demand for open voice datasets in multiple Indian languages that is openly available, accessible to everyone and of high quality. With creating and sharing open datasets, we can reduce the main barriers to voice-based technologies and create a potential market for tech innovators and social entrepreneurs.

Central Data Repository for Voice Data to Increase Access and Sharing

Multiple voice datasets in Indian languages have already been collected by researchers, start-ups or large companies. However, oftentimes they are inaccessible for local innovators because they are proprietary or simply saved locally on someone’s hard drive. A central data repository would address this problem and make open voice datasets in Indian languages readily available for everyone. The repository should bring together existing voice datasets in various languages that are currently spread across India. Newly created datasets could be added on a rolling basis. This approach would not only make voice data more accessible. It would also appreciate the diversity of datasets in terms of mode of data collection, type of recording, size, quality, domain, recording environment etc. This diversity would contribute to the formulation of standards for the Indian voice technology community.

Multi-Party Effort to Promote Open Voice Technology in India

In order to make voice datasets in Indian languages more accessible on a central repository, a multi-party effort is required. In addition to researchers at academic institutions, the private sector, civil society and international actors have an equal part to play in collecting and consolidating high-quality voice dataset. The government can facilitate this community effort of open voice data collection e.g. by encouraging the development of open-source solutions, the collection of open voice datasets and promoting the relevance of central data repositories.

Authors:

·       Dr. Kalika Bali, Principal Researcher, Microsoft Research India

·       Dr. Preethi Jyoti, Indian Institute of Technology, Mumbai

·       Dr. Prasanta Kumar Ghosh, Indian Institute of Sciences, Bangalore

·       Dr. Pushpak Bhattacharyya, Indian Institute of Technology, Mumbai

·       Dr. S. Umesh, Indian Institute of Technology, Chennai

·       Dr. Vivek Raghavan, Chief AI Evangelist, EkStep Foundation 

·       Dr. Vijay Kumar, National Language Translation Mission (NLTM), Ministry of Electronics and Information Technology (MeitY), India

·       Gaurav Sharma & Philipp Olbrich, German Development Cooperation (GIZ), FAIR Forward – Artificial Intelligence for All - Toolkit Digitalisierung