Natural Language Models and Deep Learning Models of Source Code are competing but serious roadblocks exist in terms of vocabulary size and applicability to real-world unseen projects. Worse yet, it seems that many cases language models do better than deep learned models even when the vocabulary is restricted. In this talk I will describe the work of Eddie Antonio Santos and Joshua Campbell on fixing syntax errors in a variety of programming languages using a variety of models. Furthermore the extension of treating source code as text is extended into the information retrieval domain.


Abram Hindle is an associate professor of Computing Science at the University of Alberta. His research focuses on problems relating to mining software repositories, improving software engineering-oriented information retrieval with contextual information, the impact of software maintenance on software energy consumption, and how software engineering informs computer music. He likes applying machine learning in music, art, and science. Sadly Abram has no taste in music and produces reprehensible sounding noise using his software development abilities. Abram received a PhD in computer science from the University of Waterloo, and Masters and Bachelors in Computer Science from the University of Victoria. Contact him at abram.hindle@ualberta.ca softwareprocess.ca

This page was last modified on 20 Sep 2017.