Natural Language Translation Being Implemented in Real-Time Chat Application

Abstract

For efficient machine translation, this work proposes a sequence-to-sequence model that combines Bahdanau attention with Long Short-Term Memory (LSTM) units. The encoder processes input sentences to capture contextual information, while the decoder dynamically focuses on relevant input parts, improving translation accuracy. Implemented in a real-time chat application for startups, this solution collects customer information, including preferred languages, stored in Firebase. When a customer raises a query, the owner's response is automatically translated into the preferred language, facilitating seamless communication. This approach enhances user experience and supports startups in engaging diverse clientele in global markets.