Secure & Compliant Data Management

Using Natural Language Processing To Automate Software Repairs

Secure & Compliant Data Management | Natural Language Processing | ERC

Each year, software engineers spend over half of their time testing and fixing their software. In terms of financial commitment, this means that economies around the globe allocate billions — even trillions — in resources, not to creation but instead for cleaning up mistakes. Even after launch, the fallout from software errors continues. With each update, each new piece of hardware, each relaunch, users encounter new bugs. As new reports continue to roll in, the software needs to be repaired.

Tegawende F. Bissyande (SnT)


Bug reports are one of the most important tools in the arsenal of a software developer, provide them with the information they need to replicate the conditions that lead to the error, and track down the source of the problem. While this process has long been a staple of post-release debugging and repair, it is wildly inefficient. That’s because these reports are written in natural language, with all the variety and ambiguousness that is associated with human speech. The costs, moreover, are steep: companies invest millions into repairing their software and with good reason. Software testing company Tricentis estimated that in the anglophone world alone, software malfunctions cost $1.7 trillion in lost revenue in 2017. The incentive to speed-up and improve the bug response process, and with it the software repair process, is only set to grow. In fact, the global market for bug tracking software is expected to triple by 2026.

$ 0
trillion in lost revenue in 2017

Here at SnT, our own Prof. Tegawendé F. Bissyandé has won a prestigious European Research Council (ERC) grant to research new tools that will revolutionise the way software repair — especially the bug response process — is handled. His project, NATURAL, will explore how natural language processing (NLP) technology can be used to automate the bug-response process.


In fact, NLP techniques will allow Bissyandé to create a methodology to automatically extract the fundamental meaning from bug reports. This is the first step in automating the bug-report response process. Throughout his five-year research project, Bissyandé will build on the application of NLP techniques with Machine Learning and other Artificial Intelligence tools to ultimately create a new “bot” a tool that can automatically detect and fix bugs in software, even while it is being coded.

"An ERC grant is very meaningful for the entire scientific community, as I will have the latitude to explore and discover new avenues of research during this project."

Prof. Tegawendé F. Bissyandé, SnT Tweet

Bissyandé’s ERC grant is a Starting Grant, which supports up-and-coming young researchers with stellar research track records. It provides the NATURAL project with a total of 1.5 million Euros in funding.

“The unique thing about ERC funding is that it makes it possible for me to create a team of people who will work together on the same project and build up international standing on a key research topic,” said Bissyandé. “In addition to that, an ERC grant is very meaningful for the entire scientific community, as I will have the latitude to explore and discover new avenues of research during this project. I will open doors and even if I don’t get to visit all the rooms, other researchers will then come after to explore them further.

People & Partners in this Project​

Tegawende F. Bissyande