Automatic selection of primary studies in systematic reviews with evolutionary rule-based classification

Abstract

Searching, filtering and analysing scientific literature are time-consuming tasks when performing a systematic literature review. With the rise of artificial intelligence, some steps in the review process are progressively being automated. In particular, machine learning for automatic paper selection can greatly reduce the effort required to identify relevant literature in scientific databases. We propose an evolutionary machine learning approach, called \ourmodel, to automatically determine whether a paper retrieved from a literature search process is relevant. \ourmodel builds an interpretable rule-based classifier using grammar-guided genetic programming. The use of a grammar to define the syntax and the structure of the rules allows \ourmodel to easily combine the usual textual information with other bibliometric data not considered by state-of-the-art methods. Our experiments demonstrate that it is possible to generate accurate classifiers without impairing interpretability and using configurable information sources not supported so far.