WooChat: An AI-Powered Chatbot for Academic Information Retrieval
Linda Berhe, Akhmadillo Mamirov · 2025
This paper presents WooChat, chatbot developed using Retrieval Augement Generation (RAG) based on the data from the College of Wooster website. By using Pinecone vector database and OpenAI embeddings, WooChat transforms web content into an interactive conversational system. The chatbot uses web scraping to gather data from the College of Wooster's sitemap. Gathered data is then split into document chunks and processed as vector embeddings and stored in Pinecone. We have integrated similarity search to retrieve relevant chunks of data from the Pinecone. We have implemented different ways of response generation using contextaware chain, multi-query chain and an agent-based system using GPT-3.5-turbo as our Large Language Model (LLM). This paper outlines the design, implementation, and evaluation of WooChat and discusses its potential for scalability and optimization.