Lightweight Learning Chat Bot Using Sentence Embeddings for Contextual Similarity Matching

This paper presents a minimal chat bot that uses sentence embeddings to demonstrate the core concept of semantic similarity in Natural Language Processing (NLP). Its purpose is not to compete with state-of-the-art systems in performance but to serve as an illustration showing why chat bots appear to 'understand' and where that illusion breaks under the weight of simple string-matching techniques. This paper examines the inadequacy of string-based matching for user intent recognition and introduced vector-based similarity as a conceptual framework for contextual relevance. It gives an educational bridge between naïve pattern matching and deeper semantic alignment.