Preprocessing Texts in Issue Tracking Systems to improve IR Techniques for Trace Creation

Abstract

Multiple studies showed the usefulness of requirements traceability in developing software and systems. Still, a major challenge is to establish the required trace links among development artifacts. Often, information retrieval (IR) techniques combined with text similarity measures are used for this task. Applying these ideas to requirements texts found in issue tracking systems (ITS) of open source systems is difficult, because often these texts are structured and not only contain natural language. Thus, preprocessing of the textual information is required to extract the different kinds of text. In this paper, the authors study the structure of issue descriptions found in open source systems and identify several categories of text found therein, such as source code and stack traces. These text categories allow a more precise application of similarity analysis in order to create traces by comparing textual information of the same kind, i. e. source code with source code and natural language with natural language.