Using machine learning methods to study technology-facilitated abuse: Evidence from the analysis of UK crimestoppers' text data

Abstract

Quantitative evidence on technology-facilitated abuse ("tech abuse") in intimate partner violence (IPV) contexts is lacking globally. This shortcoming creates barriers to the development of evidence-based interventions. This chapter draws on a data science-driven research project which aims to generate statistical evidence on the nature and extent of IPV tech abuse in the United Kingdom (UK). Using data from the independent UK charity Crimestoppers (2014-2019), we showcase an automated approach, facilitating Natural Language Processing and machine learning methods, to identify tech-abuse cases within large amounts of unstructured text data. The chapter offers both useful insights into the types of tech abuse found within the data, as well as the challenges and benefits computational methodologies provide. The research team has released the code and trained machine learning algorithm along with the publication of this chapter. This hopefully allows other researchers to test, deploy, and further improve the automated approach and could facilitate the analysis of other text datasets to identify tech abuse.