Investigating the identification of technical debt through code comment analysis

In order to effectively manage technical debt (TD), a set of indicators has been used by automated approaches to identify TD items. However, some debt items may not be directly identified using only metrics collected from the source code. CVM-TD is a model to support the identification of technical debt by considering the developer point of view when identifying TD through code comment analysis. In this paper, we investigate the use of CVM-TD with the purpose of characterizing factors that affect the accuracy of the identification of TD, and the most chosen patterns by participants as decisive to indicate TD items. We performed a controlled experiment investigating the accuracy of CVM-TD and the influence of English skills and developer experience factors. We also investigated if the contextualized vocabulary provided by CVM-TD points to candidate comments that are considered indicators of technical debt by participants. The results indicated that CVM-TD provided promising results considering the accuracy values. English reading skills have an impact on the TD detection process. We could not conclude that the experience level affects this process. We identified a list of the 20 most chosen patterns by participants as decisive to indicate TD items. The results motivate us continuing to explore code comments in the context of TD identification process in order to improve CVM-TD.