From User Stories to Automated Scriptless Testing via Language Models: the AUTOLINK Project

Abstract

Both requirements engineering and software testing are often regarded as tedious activities by practitioners, who tend to spend minimal effort on such tasks. Comprehensive specification documents have been replaced, in several contexts, by lightweight notations like user stories and acceptance criteria. Software tests are notoriously expensive to create and especially to maintain, leading to outdated test cases. In the AUTOLINK project, we are devising lightweight, (semi-)automated solutions that make use of NLP techniques to bridge the requirements engineering and software testing disciplines. Our NLP techniques extract a wide range of information from compact requirements representations, to then use it to automatically test software without the need of creating and maintaining test scripts. In this paper, we outline the project, describe the results we achieved so far, and sketch future directions for the project as well as for the broader research field.