Using machine learning and topic modeling to examine manifestos of violent attackers

This study is a linguistic analysis of 23 published manifestos written by violent offenders who committed or threatened a targeted attack. The acts of violence were primarily motivated by an ideology or a personal grievance that occurred between 1974 and 2022. The aim of the study was to examine if...

Full description

Saved in:
Bibliographic Details
Published in:Journal of threat assessment and management 2024-04
Main Authors: Petreca, Victor G., Burgess, Alexandra A., Wise, Jennifer, Burgess, Ann W.
Format: Article
Language:English
Subjects:
Computer Linguistics
Female
Human
Ideology
Machine Learning
Male
Simulation
Threat
Violence
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study is a linguistic analysis of 23 published manifestos written by violent offenders who committed or threatened a targeted attack. The acts of violence were primarily motivated by an ideology or a personal grievance that occurred between 1974 and 2022. The aim of the study was to examine if machine learning and the application of BERTopic was useful in forensic linguistics as it is applied to threat and risk assessment. We tested different parameter values of the techniques. Findings identified an informational table containing topic number and count, local and general topics by contrasting different settings (15 neighbors to 100 neighbors) associated with ideology, grievances, direct threats of violence, and other general topics. The semantic meaning behind planning stages, hierarchical clustering, and a heatmap of the similarity matrix showing relationships between topic shared and/or overlaps are depicted and discussed. (PsycInfo Database Record (c) 2024 APA, all rights reserved) (Source: journal abstract)
ISSN:2169-4842
2169-4850
DOI:10.1037/tam0000232