PaperFinder is an application that improves how researchers interact with scientific papers from the arXiv repository. It’s a user-friendly interface for searching, exploring, and comprehending scientific papers.

_The problem: For their work, researchers often struggle with information overload when finding and understanding relevant papers. Identifying key research findings and fully understanding their implications can be time consuming and challenging.

_Current solutions: Current tools often have steep learning curves and limited functionalities for free tier users, creating barriers to efficient research.

_The idea: Using AI to streamline paper searches, provide relevant results, and context-aware interactions to enhance understanding and comprehention.

_PaperFinder: PaperFinder is an adaptation of the original idea that leverages OpenAI API, ArXiv repository API and a user friendly interface powered by Streamlit. It offers search capabilities, AI-generated summaries and interactive Q&A features. Additionnaly, users can upload their own PDF documents which are then processed into a vector database for efficient AI-powered querying of the document’s content. With PaperFinder, researchers can reduce the time and efforts required for literature review, to focus more on analysis and original work.

_Future adaptations: Future adaptations could include custom AI models, support for multiple scientifc repositories, and better user interface.

_Learning Outcomes: This project enhanced my skills in API integration, AI application, and helped with my thesis writting. It demonstrated the potential of combining academic databases with AI to revolutionize the research process. Also, it helped me learn more about vector databases and text splitting for effective NLP.

Check out the code.

_Demo: