Analyzing Abusive Text Messages to Detect Digital Dating Abuse

Digital dating abuse is carried out using text messages, emails and social media sites and has become a major mental health crisis among the college-going population. Existing technology and non-technology based intervention programs do not provide assistance at the onset of abuse. The ultimate goal of this project is to create a mobile phone application that can detect digital dating abuse. As a step toward this goal we are investigating the use of machine learning algorithms and natural language processing techniques to flag text messages as abusive or non-abusive in the context of dating abuse. Due to the lack of a publicly available dataset that could be used to create training and testing sets for the classifiers, we first had to create and validate a dataset of abusive text messages. This paper describes the dataset creation process and the results of an evaluation of different classification and feature extraction techniques used to detect abusive texts.