Improving Systematic Literature Review Based on Text Similarity Analysis

Abstract

Systematic literature review (SLR) is an important method for identifying, evaluating, and summarizing a specific research subject. However, a SLR study always faces validity threats because there exist subjective biases in certain steps of a SLR, such as identification of relevant studies and primary study selection. In software engineering field, many systematic review studies have adopted several methods to increase the reliability, for example, group decision and pilots search. In this paper, in order to reduce subjective biases, we propose a different solution to improve objectively the quality of SLR, which aims to apply text similarity analysis into two stages of a SLR. We propose that the synonym and hypernym relations of WordNet can be used to expand the search strings to improve the paper search process. In addition, through calculating the relevance of a paper and research issues according to text similarity analysis, the primary study selection can be improved. Then, through the experiment, we validate the effectiveness of our method. Our work provides a preliminary exploration of the combination of text similarity analysis and SLR in order to improve the quality of a SLR.