Artificial Intelligence in Recordkeeping

A Systematic Review of Machine Learning Applications for Automated Records Classification

Abstract

This study presents a systematic literature review (SLR) of scholarly research published between 2014 and 2023, with the aim of identifying prevailing trends, methodological approaches, and contextual factors surrounding the use of Machine Learning (ML) models for records classification within Records Management and Archival Science. Employing the PRISMA framework, the review analyzes a curated selection of studies to assess the scope and maturity of ML applications in this domain. The findings revealed that while ML has been increasingly explored for tasks such as classification and appraisal, its application remains geographically skewed, with the majority of studies originating from Global North countries. The models employed range from probabilistic and regression-based algorithms to decision tree classifiers, reflecting diverse but largely traditional methodological approaches. The adoption of more sophisticated techniques, including deep learning and large language models, was still limited. The study underscores a critical research gap concerning the implementation of advanced ML models, particularly in the context of Global South institutions, where such technologies could significantly enhance recordkeeping efficiency and scalability. This review highlights the need for further empirical studies that develop and evaluate cutting-edge ML models in diverse archival contexts, promoting more inclusive and globally representative innovation in archival automation.