Turning text into data: Building a valuable dataset of Parole Board final decisions in England and Wales

The Parole Board for England and Wales was established in 1968 and has undergone numerous changes in its role and mandate over time. Despite its significance within the criminal justice system (CJS), it remains unresearched, with relatively limited academic engagement compared to other key areas such as sentencing. Existing research tends to be theoretical or qualitative, while quantitative work is often outdated or limited in scale. This lack of data-driven evidence contributes to persistent issues of mistrust and misunderstanding. This project seeks to create new opportunities for large-scale quantitative research on parole, arguing that such engagement shows a commitment to transparency which can foster greater trust and understanding.

One major barrier to this is the way the Parole Board currently stores information, primarily as large volumes of unstructured text in documents known as final decision letters. Natural language processing (NLP) techniques offer a valuable opportunity to extract meaningful insights from these texts. Therefore, this project explores how NLP methods can be applied to extract offending history information from final decision letters, a key piece of information in most criminal justice research. This information is then incorporated into structured datasets alongside other decision-relevant factors. The value of this data is demonstrated by using it to explore key areas of interest in parole decision-making, identified in the existing research. The result is both a new data source and a new understanding of parole decisions, contributing to greater transparency and offering a stronger foundation for future research.