Natural language processing enhanced literature reviews

Abstract

Current research manifests the challenges and limitations of analyzing copious volumes of textual content while conducting literature reviews. Recent advances in natural language processing (NLP) and increased computational power have made it easier for researchers to review large volumes of literature more efficiently. In this study, we apply Leidner’s polylithic framework as a theoretical lens to examine areas where NLP may aid the literature review process. Specifically, we anatomize the use of NLP in published literature review studies in terms of capabilities, characteristics, and workflows. Through this analysis, we identify four meta-requirements that underpin the development of effective NLP-enhanced literature review solutions, forming the foundation of our proposed NLP-enhanced Literature Review (NLP-e-LR) framework. This framework provides structured guidance for applying suitable NLP techniques across various types of literature reviews based on research objectives and review focus. We present a mapping of NLP-e-LR process enhancements to the core literature review tasks (i.e., search, screening, and analysis), outlining a range of NLP capabilities aligned with each task and detailing their relevance across different literature review types. Through illustrative examples, we demonstrate how NLP-e-LR facilitates key literature review activities. Finally, we discuss the benefits and limitations of the framework and identify directions for future research.