Inspiration
Tired of having to read long texts (i.e. your lecture notes) that are boring and difficult to understand? We get you, manually taking notes of the key points and clustering them together can be a tedious process. As research shows that visual learning enables one to understand and retain information better, our project aims to create an engaging and interactive knowledge graph from selected text to aid users to understand the main concepts in a matter of seconds.
What it does
Knowledge.Ai consists of a browser extension and web application with a NLP pipeline which accepts text input (either select text from websites or copy paste onto the web application) and extracts relations between entities to form a knowledge graph. This allows for a cleaner and easy-to-understand visual representation of the long texts.
How we built it
We built our browser extension using react and typescript and web application using react and django with a mysql database. Our AI model for relation extraction is a pretrained model from huggingface where the model outputs are visualized using a react library for knowledge graph generation.
Challenges we ran into
- Sourcing and creating an interactive knowledge graph
- Developing an extension with react and typescript
- Setting up an extension and making it communicate with our AI model
- Debugging code for user authentication
Accomplishments that we're proud of
- Getting the extension up and working as we had no experience working with extensions before.
- Integrating persistent login into the web application
- Integrating authorization tokens for user authentication and registration
- Being able to display an interactive graph of key points and their relations from text input
What we learned
We learned how to create an extension with react and typescript and how make it communicate with an AI model. We also learned that we should never give up when facing an error/bug. Eventually we will be able to solve the issue even when it takes a lot of time. We also learned that we were able to work without any sleep for 24 hours.
What's next for Knowledge.Ai
Different enterprise could develop different use cases, depending on the data that they are able to consolidate. For example, we can fine-tune the model on a financial dataset, which can help speed up the preliminary research of an prospective investment opportunity. We also plan to expand the use cases to the technical research domain, where we fine-tune the model on research documents. This enables researchers to review the key points of each paper while comparing them, making the research process more efficient.
Log in or sign up for Devpost to join the conversation.