Idea: an AI system using OpenAI API to help process primary source research documents to make preliminary research easier for research projects
Features:
- Be able to provide information on a specific research topic given a query
- Provide information on specific research papers given the link
Steps in pipeline:
- First fetch the text given by the query and find the first top matching research documents
- Extract the text
- For research paper summary: run a total summary or a summary on each section of the paper
- For general topic summary: fetch the documents for the query and run a summary on the abstracts and methods
Technologies Required:
- Selenium - for browser automation searching
- Frontend: Next.js
- Selenium worker: go
- OpenAI API: express or fastapi
OpenAI API Prompts:
- Best way to set up as fetching the link, sending the text to OpenAI (chunking if required or separate sections)
- When checking for general topic: send text and ask for summary in each (bullet point based)
- Requires some experimentation with the prompts