Assessing the Effectiveness and Efficiency of the Offense Text Auto-Coder

• In Event: Novel Uses and Approaches to Administrative Data

Abstract

Criminal justice agencies throughout the United States collect information using free-text fields, which permits a person entering data to record unique character responses despite the existence of discrete categories. This is nowhere more evident than in court records, where discrete offense information – e.g., murder – can be entered into data collection systems in numerous ways – mrder, murdr, muder, etc. Although the existence of free-text fields eases the process of entering information on the front end, the numerous iterations of characters that can correspond to a single offense can create substantive delays during data processing and analysis. As such, the current study developed the Offense Text Auto-Coder (OTAC) which was designed to ease the time burden associated with manually processing free-text offense information received from court, jail, and pretrial service agencies across the United States. In particular, OTAC is a machine learning algorithm designed to semi-automate the process of transforming free-text offense information into National Corrections Reporting Program (NCRP) codes. The current study evaluated the effectiveness and the efficiency of OTAC, with a focus on assessing the potential use cases for the tool outside of the current study. The benefits of OTAC for processing data will be reviewed during the presentation.

Authors

• Matthew DeMichele, RTI International

• Ian A. Silver, RTI International

• Peter Baumgartner, RTI International

• Jason Walker, RTI International