Programming with Big Code

Informations projet

AutoCode

N° de convention de subvention: 727579

DOI

10.3030/727579

Projet clôturé

Date de signature de la CE 27 Avril 2017

Date de début 1 Mai 2017

Date de fin 31 Octobre 2018

Financé au titre de

• EXCELLENT SCIENCE - European Research Council (ERC)

Coût total

€ 150 000,00

Contribution de l’UE

€ 150 000,00

Coordonné par

TECHNION - ISRAEL INSTITUTE OF TECHNOLOGY
Israel

Periodic Reporting for period 1 - AutoCode (Programming with Big Code)

Période du rapport: 2017-05-01 au 2018-10-31

Résumé du contexte et des objectifs généraux du projet

The goal of AutoCode is to enable automatic programming based on machine-learning over large code bases.
In this project, we developed several interesting prototypes that represent significant progress in the area of automatic programming.
One of our protoypes, called code2vec, is available at www.code2vec.org and is able to predict labels that describe a given piece of code.
The system has been released as open-source (https://github.com/tech-srl/code2vec).
Another prototype that we developed, called code2seq, is available at www.code2seq.org and is able to predict a natural language sentence that describes a given piece of code.
The system will be released as open source under https://github.com/tech-srl.
Both systems have gained much interest both in academia and in industry.
There are several applications of the technology in Industry – at Microsoft Research, in VISA research, and in various startups.