Paper review: “Learning to Represent Programs with Graphs”

ML on Code is a rapidly developing field, both in academia and industry, that source{d} was set out to systematically explore throughout the last years. So far the results published by our Data Retrieval, Machine Learning, and Infrastructure teams who collect and store millions of Git repositories were based on large-scale applications of advanced NLP techniques such as: Identifiers Embedding, Topic Modeling and sequence model for Identifier Splitting. Current research avenues, driven by the applications for assisted code review, include models using more structured representations of the source code, based on Universal Abstract Syntax Trees and graphs.

This is a companion discussion topic for the original entry at