Learning Embeddings for Users, Ads, and Tweets based on the Twitter Engagement Graph

What is it about?

Individuals on Twitter author Tweets as well as engage with them (favorite, retweet, reply). In this work we take these engagements and create a large Twitter network of engagements between users, tweets, and ads. We then learn an embedding of nodes. We demonstrate these nodes can be used for a variety of machine learning and personalization tasks.

Featured Image

Why is it important?

We show a simple embedding method can scale to gigantic graphs. Additionally we demonstrate that such graph embeddings can be invaluable to industry recommendation tasks.