Inspiration
We needed a graph problem. Citation graphs are cool! The dataset we used has keyword features for each node which help with base classification.
What it does
Proof of concept model that uses Graph Deep Learning to weigh node parameters based on neighbouring node values. It gives a prediction based on an academic research papers text and its neighbouring cited papers. Using a graph based model improves accuracy by up to 70% on a 7-class classification.
How we built it
Python, and basic NumPy/Tensorflow in a Jupyter Notebook
Challenges we ran into
We really wanted to get a front-end working where users could type in a list of citations and find links between them, but it was harder then we thought, especially when generating feature vectors that matched the dataset and our trained model.
Accomplishments that we're proud of
Learning Graph Neural Networks
What we learned
Basics of Graph Neural Networks!
What's next for AcademAI
See Challenges
Log in or sign up for Devpost to join the conversation.