In a systematic literature review, data is collected from what others have published and is synthesized to give an overview of a subject area. It can, for example, be used to see which software engineering processes that are most commonly used to solve a particular problem or the pros and cons of using a particular software architecture in a specific type of systems. A systematic literature review is referred to as a secondary study, meaning that no firsthand data is collected. Only data published by others is used.
In a systematic literature review, it is important to have a structured way of searching for articles. You need to define where you search and what search terms that are used. An example is that you search for the term “sorting algorithm performance” in the ACM and IEEE Xplore Digital Libraries.
The search will result in a list of articles. If you have well-defined search terms most of the articles in the list will be relevant for your study, but some will most likely not be relevant. To combat this you define some inclusion-exclusion criteria and read the abstract of every article to check if it matches your criteria. An example is to only include articles that mention bubble sort, quick sort, selection sort, insertion sort or merge sort in the abstract.
To find good search terms and inclusion-exclusion criteria you often have to start with one or more trial searches. The purpose of these is to try different search terms and criteria since it can be tricky to directly come up with the perfect search terms.
In a systematic literature review it is important that you document and report everything you do (often referred to as your search strategy):
- Search terms used
- Which libraries and search engines that were used for the search
- If you exclude old articles (published before a specific year)
- Inclusion-exclusion criteria used
- When you performed the search (date)
A complete guide can be found here. It is recommended that you read it before planning and conducting your own systematic literature review.
