HMwaniki-stack/NLP-Chatbot

This project aims to develop a natural language chatbot capable of generating human-like responses and understanding informal customer feedback expressed in English, Kenyan Swahili and Sheng. # NLP-Chatbot This project aims to develop a natural language chatbot capable of generating human-like responses and understanding informal customer feedback expressed in English, Kenyan Swahili and Sheng. ## 1.1 Problem Statement Startups entering new markets often struggle to understand customer feedback when it's expressed in local dialects or informal language. In Kenya, much of this communication occurs in Swahili and Sheng, which combine local slang, English, and Swahili in a fluid, often unstructured manner. Existing chatbot systems trained on formal English fail to grasp the tone, intent, or meaning behind such messages. This project aims to fill that gap by building a chatbot trained specifically on real-world Kenyan conversations to interpret and respond to customer queries and feedback with local context and relevance. ## 1.2 Objectives - Collect and preprocess Kenyan user dialogue from YouTube, social media, and local content featuring Swahili and Sheng - Fine-tune the chatbot with foundational data for conversational structure, while emphasizing local language patterns - Build a sequence-to-sequence model capable of handling informal, code-switched dialogue - Evaluate the chatbot’s performance with emphasis on contextual relevance and local understanding - Present a working prototype that simulates real customer feedback scenarios