Google recently introduced its open-source AI model, Gemma, which has seen widespread adoption since its launch a few months ago. At the annual developer conference, Google I/O 2024, developers showcased the use of Gemma in creating diverse projects, including Navarasa, a multilingual AI variant designed for Indian languages.
Project Navarasa, developed by Telugu LLM Labs, utilizes Gemma to make AI accessible in 15 Indian languages. The project aims to bridge the language barrier by enabling AI-driven language generation across different Indian languages. Gemma’s powerful tokeniser plays a key role in this endeavor, allowing the model to adapt to various alphabets and language systems.
Harsh Dhand, head of APAC research partnerships at Google, emphasized the significance of developing technology that caters to India’s cultural nuances. He highlighted the importance of building models like Navarasa to facilitate communication in native languages, thereby enhancing user experience.
The developers behind Navarasa expressed their vision of creating culturally rooted language models that enable users to interact with chatbots in their preferred language. By incorporating a wide vocabulary of words, symbols, and characters, Navarasa aims to cater to diverse linguistic preferences across India.
In comparison to other models like Llama, several developers have lauded Gemma’s efficiency in working with Indian languages. The versatility and adaptability of Gemma have positioned it as a valuable tool for language-related projects like Navarasa.
Overall, Project Navarasa exemplifies Google’s commitment to promoting linguistic diversity and inclusivity in the realm of AI-driven technologies. With the integration of Gemma’s advanced capabilities, Navarasa paves the way for enhanced communication experiences in multiple Indian languages, driving innovation and accessibility in the digital space.
