Measuring interpersonal firearm violence: natural language processing methods to address limitations in criminal...

Area of Science:

• Public Health
• Computer Science
• Criminology

Background:

• Firearm violence is a significant public health issue in the U.S.
• Existing data infrastructure is insufficient for comprehensive study.
• Court records contain valuable information on violent crimes.

Purpose of the Study:

• To develop and evaluate a natural language processing (NLP) method for classifying court documents.
• To identify firearm involvement in alleged violent crimes using NLP.
• To assess the feasibility and accuracy of NLP for this task.

Main Methods:

• Digitized and reviewed 1472 court records from Washington state.
• Developed a key term search and trained supervised machine learning classifiers.
• Evaluated classifier performance using a held-out test set.

Main Results:

• A decision tree classifier achieved the best performance with an F1 score of 0.82.
• A key term list demonstrated perfect recall (1.0) but a lower F1 score (0.65).

Conclusions:

• NLP is an accurate and feasible method for identifying firearm involvement in court documents.
• This approach has the potential to save significant time in data analysis.
• The findings support the use of NLP for studying firearm violence.