AI for Record Appraisal

Traditional approaches to reviewing massive volumes of born-digital records are not viable. It is widely accepted that the appraisal, selection, and sensitivity review of digital records will only be feasible with machine assistance.

This applied research project aims to leverage Large Language Models (LLMs) to provide insights into large volumes of information to support record managers’ appraisal decisions. We are exploring techniques such as transforming unstructured text into structured forms enable auto-labelling for classification and clustering. Our focus is on approaches like Retrieval-Augmented Generation (RAG) and topic modelling to enhance context understanding, enable information extraction, and facilitate context-based analysis of digital records. The experiments also aim to explore if LLMs’ capabilities can enable techniques like taxonomy generation and semantic analysis, which are proven concepts but were previously too complex and expensive to implement at scale.