Building Content Representations at Twitter with Self-Supervised Learning

For the last few quarters, I have been working at the intersection of Recommendations and NLP, thinking about ways to represent textual information for Twitter’s recommendation products. Recently, my colleague Kenny Leung and I gave a talk at ODSC 2020 about our experience building and serving self-supervised content representations for recommendations on Twitter.

The slides for the talk can be found here.

self-supervised-training