Inspiration
As a former master's student in science, I know the difficulties of trying to contribute to science. Today, science is so advanced that the requirement to create original research for our thesis seems impossible: with thousands of articles available, we feel overwhelmed by the amount of information, but ironically, it's difficult to find research that helps us. It's not uncommon for a student to spend more than half their time reading thousands of articles without knowing if they will be relevant to future research.
What it does
Paper Pal AI: Your AI-Powered Research Partner for Deeper Insights transforms the way you engage with academic literature, moving beyond traditional search to provide a truly intelligent and interactive research experience.
Here's how Paper Pal AI revolutionizes your research workflow:
- Natural Language Discovery: Forget endless keyword combinations. Simply describe your research interest in natural language, and our system uses semantic search to instantly find the most relevant articles, even if they don't contain your exact terms.
- Intelligent Summaries & Key Takeaways: Dive straight into the core. For any article, get immediate, AI-generated concise summaries and actionable bullet-point key takeaways, saving you countless hours of reading and accelerating your comprehension.
- Dynamic Recommendations & Refinement: Receive personalized recommendations that truly matter. Browse your recommended lists, and if an article isn't relevant, simply delete it to refine future suggestions. Found a gem? Get more relevant articles based directly on that one, expanding your insights with precision.
- Uncover Relevance: Ever wonder why a particular article showed up? Just ask why it's relevant, and our AI will provide a clear explanation, enhancing your understanding of the connections within the literature.
- Spark New Research Directions: The innovation doesn't stop at discovery. Take a curated list of articles and let Paper Pal AI help you brainstorm new research directions. Our system will analyze the collective themes and ideas to suggest novel avenues for your next big project.
How we built it
In this project, we leveraged Google Cloud's advanced artificial intelligence (including Gemini and text embedding) and MongoDB Atlas Vector Search.
Using the ArXiV dataset and Google Cloud's document retrieval services, we created a MongoDB Atlas database with all the information needed for our project. The backend implements basic functionality with vector search and content generation using both services, all implemented in FastAPI. Meanwhile, the UI was implemented using React, which, although simple, provides all the desired functionality.
What's next for Paper Pal AI
This project precisely utilizes mongdb's vector search capabilities, but new filters could also be implemented, such as combining vector search with filtering by author or arxiv tags.
We could also improve response speed by using caching systems, thus avoiding the use of the Gemini API as much as possible, thus reducing costs.
But these are improvements that could easily be implemented with more time.
Log in or sign up for Devpost to join the conversation.