Tool support for systematic reviews in software engineering

Abstract

Background: Systematic reviews have become an established methodology in software engineering. However, they are labour intensive, error prone and time consuming. These and other challenges have led to the development of tools to support the process. However, there is limited evidence about their usefulness.

Aim: To investigate the usefulness of tools to support systematic reviews in software engineering and develop an evaluation framework for an overall support tool.

Method: A literature review, taking the form of a mapping study, was undertaken to identify and classify tools supporting systematic reviews in software engineering. Motivated by its results, a feature analysis was performed to independently compare and evaluate a selection of tools which aimed to support the whole systematic review process. An initial version of an evaluation framework was developed to carry out the feature analysis and later refined based on its results. To obtain a deeper understanding of the technology, a survey was undertaken to explore systematic review tools in other domains. Semi-structured interviews with researchers in healthcare and social science were carried out. Quantitative and qualitative data was collected, analysed and used to further refine the framework.

Results: The literature review showed an encouraging growth of tools to support systematic reviews in software engineering, although many had received limited evaluation. The feature analysis provided new insight into the usefulness of tools, determined the strongest and weakest candidate and established the feasibility of an evaluation framework. The survey provided knowledge about tools used in other domains, which helped further refine the framework.

Conclusions: Tools to support systematic reviews in software engineering are still immature. Their potential, however, remains high and it is anticipated that the need for tools within the community will increase. The evaluation framework presented aims to support the future development, assessment and selection of appropriate tools.