Today the University announced that Prof. Tegawendé F. Bissyandé from the University of Luxembourg’s Interdisciplinary Centre for Security, Reliability and Trust (SnT) has been awarded an ERC Starting Grant to execute research in software repair.
We sat down with Bissyandé to get more details about the story and what his research path is.
How did you arrive at this research question?
During my PhD I researched tool support for debugging operating system drivers. However, this was only about identifying fault locations. My dream, and that of the whole community, has always been to go further and fix software bugs automatically. During a PhD internship at Singapore Management University (SMU), I had the opportunity to explore data mining techniques to improve software maintenance work. This is where I had the idea of automatically mining bug reports to repair software. I have had this idea in my head for many years and now it is finally the right time to begin.
Why is it the right time?
The project is timely for the simple reason that the technology needed is ready now. Natural language processing has advanced a lot, and deep learning is showing very promising results, so all the tools we need to work with are there. They weren’t five years ago.
What is your objective?
This ERC grant will have a significant impact as it tries in its core to reconcile research excitement with practitioners’ expectations and constraints. Today automatic programme repair seems merely just a compelling area of scientific research, and I want to transform it into something that can be used in the real world. We shouldn’t just build things in research environments that can never be applied in practice. This is a key focus for all the research projects at SnT, and our industry partners will make it possible for me to assess the viability of our research outcomes on this project.
How will you get there?
My research path for this project is defined by four milestones.
- First, we need to develop programme-related natural language processing further so that it can be used to truly understand the end user’s issues with a piece of software when the technology is deployed as a bot.
- Once we have achieved that, we then need to make it possible for the bot to communicate with the programme developer, so that they can understand the issue in their own professional context. As a simple example, a user would normally input a complaint that says something like “I press this button to do XX and it doesn’t work”, which the bot could then understand, investigate and go back to the developer with a report that would be something like this “there is an error in the 158th line of code that is causing XX malfunction”. This would already save developers a significant amount of time, but I intend to go even further.
- As a third milestone I will research how to make it possible for the bot to be able to build a test suite automatically, so that the developer can plug-in their solution right away and see if they have solved the problem.
- The final phase for the research will be to use these capabilities altogether to make it possible for the bot to autocorrect code as a developer is writing it. This would be very similar to the way that your phone autocorrects your texts as you type them.
When did you submit your application?
I submitted in October, and between then and now is an extensive process that includes two review phases that assess both the potential of the candidate and the scientific quality of the proposal. There are thousands of applications from very highly-qualified researchers. However, in the end, only about 1 in 10 applications are accepted.
What role did SnT play in your application process?
I am quite conscious that I might not have been successful applying for an ERC Starting grant if I had been a researcher anywhere else. SnT has been the most supportive environment I could imagine as a researcher. Since arriving here in 2013 as a post-doctoral researcher I had the opportunity to advise PhD students right away and was put in a position of framing research questions from the very beginning. As a post-doc anywhere else you are usually expected to follow a rigid agenda and don’t have as much room for growth. SnT was undoubtedly the best place to kick-start my career, and I am very pleased that I have been able to stay here, now as the deputy head of a research group.