Surveillance of Domestic Violence Using Text Mining Outputs From Australian Police Records

• DOI: 10.3389/fpsyt.2021.787792
• Publication Date: 2022-02-09T05:36:29Z
• Authors: George Karystianis; Armita Adily; Peter W. Schofield; Handan Wand; Wilson Lukmanjaya; Iain Buchan; Goran Nenadic; Tony Butler
• Keywords: Depression

In Australia, domestic violence reports are mostly based on data from the police, courts, hospitals, and ad hoc surveys. However, gaps exist in reporting information such as victim injuries, mental health status abuse types. The police record details of events structured (e.g., gender, postcode, ethnicity), but also text narratives describing other substance use, status. voluminous nature has prevented their use for surveillance purposes. We used a validated mining methodology 492,393 police-attended event 2005 to 2016 extract mentions persons interest (POIs) (individuals suspected/charged with offense) victims, types, injuries. A significant increase was observed that recorded an injury type (28.3% 35.6% 2016). pattern types differed between male female victims more likely be punched experience cuts bleeding grabbed pushed have bruises. four most common illnesses (alcohol abuse, bipolar disorder, depression schizophrenia) were same POIs. An 5.0% 24.3% proportion reported illness among victims. These findings demonstrate extracting can provide novel insights into patterns including confounding factors illness) thus enable policy responses address this public problem.