Harpreet Singh Anand | The Benchmarking Conference | LDC-IL

Multilingual Indic Real-Time Chat System

Harpreet Singh Anand

Birla Institute of Technology and Science,
Pilani - Pilani Campus, India


Authors : Harpreet Singh Anand, Amulya Ratna Dash, Yashvardhan Sharma

Abstract

In today's globalized world, language barriers often hinder effective communication. We present a real-time multilingual chat system that enables seamless conversations between users speaking different languages. Built using Streamlit and integrated with Google Translator, NLLB and open-source models, the system ensures that messages are instantly translated into each user's preferred language. Unlike conventional translation tools, our system maintains a natural flow of conversation by dynamically translating messages in real time, allowing users to interact effortlessly in their native languages.

Our application features a user-friendly interface where participants can join a chat room, select their preferred language, and exchange messages with automatic translation. Our system uses cosine similarity between the input sentence and the generated translations(from various models) for the translation quality evaluation and outputs the most accurate translation.To enhance usability, the system employs an optimized backend for storing and retrieving chat history, ensuring that conversations persist across sessions. Additionally, real-time updates allow users to see translated messages instantly without manual refreshing. This application has wide-ranging applications in cross-cultural collaboration, education, and international business, making communication more inclusive and accessible.