BookQuery

A project built to chat with PDF documents.

You have a PDF document you want to chat with in other to get insights, this tool is a Jupyter Notebook to consider as a Proof-Of-Concept using the following process:

• Extract text from PDF using langchain.document_loaders UnstructuredPDFLoader,
• Chunk the result data up into smaller documents using langchain.text_splitter RecursiveCharacterTextSplitter,
• Create embeddings of the documents using Open AI Embeddings,
• Make an indexation of the documents embeddings and store them into a pinecone database,
• Query those documents.

For each user query, we process it as following:

• Create embedding of the user query using Open AI Embeddings,
• Make a semantic search in the indexed embeddings database,
• Make a summary of the result using Langchain question answering and OpenAI GPT 3.5 Turbo model. This summary is the chatbot answer.