LISAN is a research driven, AI powered chatbot designed to understand and generate
responses in multiple Arabic dialects. Built as part of our Senior Project at the
University of Sharjah, LISAN addresses a major limitation in todays Large Language
Models (LLMs): their struggle to naturally process the diverse dialects spoken across
the MENA region.
Our system uses a multi agent LLM ensemble that detects the users dialect
(such as Gulf, Levantine, North African, Iraqi, Yemeni, Nile, or Modern Standard Arabic)
and responds in the same dialect. Through extensive testing, weighted voting, and backend
integration using n8n, LISAN delivers a smooth, culturally adaptive conversation experience
that stands apart from existing single dialect or MSA only systems.
LISAN represents our ambition to create inclusive, accessible, and dialect aware AI tools
for Arabic speaking communities.