BHASHINI and IISc Bangaluru : Text-to-Speech Model for Sourashtra Language

Key Highlights:

• The Digital India BHASHINI Division (DIBD) under MeitY and IISc Bengaluru have developed Text-to-Speech (TTS) models for the Sourashtra language.
• Sourashtra is a minority Indo-Aryan language predominantly spoken in Tamil Nadu, identified as a low-resource language.
• The models utilize the VITS architecture and generate speech from text, supporting both Tamil and Sourashtra scripts with male and female voices.
• A dedicated Sourashtra speech corpus (over 28 hours) developed at IISc's SPIRE Lab was used for training and is available on BHASHINI's AIKOSH platform.
• This initiative aims to enhance digital inclusion, language learning, and accessibility for the Sourashtra-speaking community.

Detailed Insights:

• The project addresses the challenge of limited digital resources and speech technology for underrepresented Indian languages.
• The VITS architecture allows direct character-level text processing, eliminating the need for a separate phonemizer.
• The corpus includes diverse content from literature, with transcriptions available in both Sourashtra and Tamil scripts.
• The models can support applications like reading Sourashtra text aloud, language learning tools, and accessibility solutions.
• BHASHINI aims to build a national ecosystem for AI-powered multilingual digital inclusion, supporting numerous Indian and international languages.
• This development contributes to making digital information and services more accessible across India's linguistic diversity.

Key Concepts Involved:

• Text-to-Speech (TTS): Technology that converts written text into spoken audio.
• Low-resource languages: Languages with limited digital data, tools, or computational resources for natural language processing.
• VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech): A specific neural network architecture used for high-quality speech synthesis.
• Digital India BHASHINI Division (DIBD): A government initiative under MeitY to build an AI-powered multilingual digital inclusion ecosystem.
• AIKOSH: BHASHINI's platform providing access to language datasets and resources for Indian language technologies.