Case study

How the Ministry of Justice used AI to compare prison reports

Learn how the Ministry of Justice (MoJ) used natural language processing to identify patterns across prison reports.

This guidance is part of a wider collection about using artificial intelligence (AI) in the public sector.

AI application used

  • natural language processing (NLP)

Objective

MoJ needed to compare how various factors including geography, and incidents such as inmate conflict, affected different prisons.

Situation

MoJ had over 250,000 sentences of unstructured text in over 500 reports detailing their inspections of prisons and other institutions. These reports were from:

  • HM Inspectorate of Prisons
  • the Independent Monitoring Board
  • the Prisons and Probation Ombudsman
  • Ofsted reports into secure training centres

There were too many reports for staff to quickly access relevant information.

Action

MoJ trained a neural network on the prison reports to track how people use specific words in prison contexts. The algorithm groups words with similar meanings to form an ‘intelligent search’ tool. New reports are automatically added to the tool’s library so the data remains up-to-date. This means staff can rapidly uncover information buried in the reports and identify trends.

Impact

The tool helps MoJ:

  • identify patterns of issues and incidents across prisons
  • identify geographic patterns affecting prisons
  • inform data-driven decisions about prison inspections and policy

Updates to this page

Published 26 June 2019