This body of work focuses on the design and implementation of an automated text-mining method that extracted information from a large-scale set of family and domestic violence (FDV) police records (n=492,393) and on whether this information can be used for further research. We report an evaluation of the method and the results from its application to the large dataset. These include trends over time for mentions of mental health problems obtained from police records for individuals involved in FDV events; specificity and sensitivity when comparing the extracted mental disorder mentions with NSW Health data; and the development of a predictive analytics approach to breaches of apprehended violence orders based on the extracted information. Findings indicate not only that text mining the free-text FDV police records can yield substantially useful and previously unknown information but also that text mining can fuel predictive analytics that can indicate high-risk offenders in the FDV area, impacting early prevention and intervention policies in FDV cases.