Programming with Big Code

Información del proyecto

AutoCode

Identificador del acuerdo de subvención: 727579

DOI

10.3030/727579

Proyecto cerrado

Fecha de la firma de la CE 27 Abril 2017

Fecha de inicio 1 Mayo 2017

Fecha de finalización 31 Octubre 2018

Financiado con arreglo a

• EXCELLENT SCIENCE - European Research Council (ERC)

Coste total

€ 150 000,00

Aportación de la UE

€ 150 000,00

Coordinado por

TECHNION - ISRAEL INSTITUTE OF TECHNOLOGY
Israel

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

Período documentado: 2017-05-01 hasta 2018-10-31

Resumen del contexto y de los objetivos generales del proyecto

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.