Multilingual AI chatbot using transformers

The rapid advancement of Natural Language Processing (NLP) has enabled the development of intelligent conversational agents capable of interacting in multiple languages. This paper presents a scalable and efficient multilingual AI chatbot architecture that integrates transformer-based language detection with large language model (LLM)-driven response generation. The proposed system employs a Bidirectional Encoder Representations from Transformers (BERT) model for accurate language identification, followed by a Retrieval-Augmented Generation (RAG) pipeline powered by LLaMA 3 via Groq for response generation. To enhance contextual relevance, LangChain is utilized for orchestration, while Qdrant serves as the vector database for semantic retrieval. Mistral-based embedding models are used to convert textual data into dense vector representations, enabling efficient similarity search across multilingual corpora. The frontend is developed using React and Tailwind CSS, while the backend leverages Python for model integration and API handling.
The system aims to provide accurate, context-aware, and language-specific responses across diverse linguistic inputs. Experimental observations suggest that combining transformer-based language detection with modern LLMs significantly improves chatbot performance in multilingual environments. This architecture is particularly suitable for real-world applications such as customer support, education, and cross-lingual communication systems.