3 days ago
Students' AI project aimed at preserving Marathi literature
PUNE: In a remarkable initiative blending artificial intelligence (AI) and regional heritage, four final-year computer engineering students from the Pimpri-Chinchwad College of Engineering (PCCOE) have developed 'Marathi Bhasha Vishwa', an innovative AI project aimed at preserving and promoting the Marathi language. The project features two unique systems namely 'Marathi Tiny Stories' and 'Sant Sahitya AI' that marry the power of small language models (SLMs) with the richness of Marathi literature.
About the motivation behind the project, student Anish Joshi, said, 'We realised that while there is a lot of focus on AI for global languages, regional languages like Marathi are often ignored. With this project, we wanted to ensure our language finds a voice in the digital world too.'
The 'Marathi Tiny Stories' system uses a fine-tuned BLOOM-560M model to generate short, culturally aligned stories in Marathi for children that are ideal for rural schools and digital classrooms with limited infrastructure. The stories are created using carefully designed prompts and include titles, dialogues, and morals in simple language appropriate for children aged three to five years.
Aryan Sagavekar, who worked on the story dataset generation, noted, 'We designed each story to be emotionally resonant and educational. It was important to us that the stories reflect Marathi culture and also instil values in children.'
The second system, 'Sant Sahitya AI' is designed for philosophical interpretation. Users can upload texts from saints like Sant Dnyaneshwar and Sant Tukaram and the system provides explanations and insights into the tone and literary style of the original works.
Guruprasad Pathak, who worked on this module, explained, 'Sant literature is deep and often hard to interpret. With AI, we created a way for modern readers to understand the essence of this literature without losing its poetic soul.'
The team ensured that the models were optimised for low-resource environments, employing techniques like prompt engineering, fine-tuning, and creative dataset design. The models can run on mobile devices or basic computers, making them accessible across urban and rural regions.
Anish Joshirao, summarising the team's vision, shared, 'This is more than just a project. It is a tribute to our language. We hope it inspires others to build similar tools for regional languages across India.'
The team attributed their success to the guidance of their mentor, professor Rachana Patil, head of the computer engineering (regional language) department and support from the PCCOE's department of computer engineering. According to the students, future work may include voice interfaces, story visualisations, and expanding the platform to other Indian languages.
This student-led project stands out as a shining example of how technology can empower cultural preservation, proving that AI, when applied thoughtfully, can become a powerful force for linguistic inclusivity and heritage education.
About this project, professor Patil said, 'This project is a beautiful confluence of technology and tradition. What impressed me the most was the students' sensitivity toward cultural heritage and their commitment to using AI not just as a tool, but as a bridge to connect generations. 'Marathi Bhasha Vishwa' goes beyond coding and algorithms; it reflects a deep understanding of language, ethics, and accessibility. The students were incredibly dedicated, from curating culturally authentic datasets to ensuring that the models run smoothly even on low-end devices. As a mentor, seeing them apply advanced AI techniques to preserve and democratise Marathi literature has been truly fulfilling. I believe this project sets a benchmark for how regional languages can find meaningful space in the AI ecosystem.'
