The University of Oviedo's Computational Reflection research group has worked with Cambridge University to develop a system that uses big data to enable large volumes of source code (the lines of text that contain the steps to be followed by a computer when running a program) to be processed in seconds. To do this, the software is represented not as text (source code) but instead as superimposed graph structures that represent the code's semantics and syntax.

The system that has been created is capable of picking up complex programming errors within Java projects containing millions of lines of code in just tens of seconds, where existing systems fail due to consuming too many computer resources. The system also allows highly useful software properties to be extracted in order to create artificial intelligence models. As an example, the research team has applied machine learning to estimate the level of experience of programmers simply by analysing their source code.

This project has been funded by Spain's Ministry of Science and Innovation. The results have been published in the scientific journals IEEE Access and Future Generation Computer Systems. The project forms part of Óscar Rodríguez Prieto's doctoral thesis, directed by Francisco Ortín, to be presented on Friday 12 June.