Constrained Creative Generation

Computational generation of poetry in Indian languages

Poetry in Indian languages is bound by strict rules of metre (chandas), alliteration, and tradition — constraints that neural language models have no reason to respect unless explicitly guided. This project builds generation systems and evaluation methods that take these constraints seriously, going beyond fluency to poetic form and cultural appropriateness.

Team: Asmit Mukherjee

External Collaborators: Teams from IITKGP, IITK