The idea is to be able to build a full fledged question answering system that can condense knowledge from uploaded pdf documents well enough to be able to answer a set of questions.
An android application which displays 'parts of speech' tags for questions entered using Apache's OpenNLP library(https://opennlp.apache.org/) .
Demo
Divide this into smaller cycles , a couple of objectives would be.
- User is able to upload any pdf document.
- Introduce page limits if needed or assess how hard it is to upload pdfs.
- develop the entire interface to be able to upload a pdf in the application.
-
Simple question answering system on android to handle easy questions based on the knowledge that can be derived from the uploaded documents.
-
Refine the question answering system.
- Access the need for a knowledge graph or knowledge base of any kind.
- Specific section under references for refining question answering systems.
- Enter LLMs - develop a streamlit RAG app that allows users to upload documents and ask questions Done
- Conversational question answering for alexa by amazon - https://www.amazon.science/publications/knowledge-informed-semantic-parsing-for-conversational-question-answering