Text mining domestic violence police event narratives

The advent of big data offers several opportunities for knowledge discovery enabling researchers through the use of automated methods to process large datasets in a relatively small amount of time. The NSW Police Force attends each year thousands of family and domestic violence (FDV) events recording a wealth of information in unstructured text narratives including mental illnesses for victims and perpetrators, occurred abuse types, sustained injuries and various threats for harm. Yet, this unique source of information remains widely untapped for research purposes. In a world’s first, we designed, evaluated and implemented a text mining methodology in a large dataset of police FDV narratives (493,292) to identify automatically key information for individuals involved in FDV. Our research demonstrates that text mining police narratives is not only feasible, but it can lead to unique insights for FDV surveillance at a population level. The extracted information can be used successfully as an input towards state-of-the-art machine learning models that predict future offences within the area of FDV and assist in shaping early prevention and intervention policies.