tamuGPT
2024
tamuGPT is a retrieval-augmented chatbot for answering questions over indexed web content. It began with Texas A&M University sources before evolving into a general-purpose question-answering system.
The system crawls websites and applies custom preprocessing and summarization to the collected documents. Its retrieval pipeline combines vector search, word2vec, and BM25 to balance semantic similarity with keyword coverage.
The retrieved passages are reranked before the most relevant context is supplied to the language model, improving the relevance of generated answers.
Technologies
Python, LangChain, GPT, Scrapy, Pinecone, word2vec, and BM25.