Intent Classification using Feature Sets for Domestic Violence Discourse on Social Media

Domestic Violence against women is now recognized to be a serious and widespread problem worldwide. Domestic Violence and Abuse is at the root of so many issues in society and considered as the societal tabooed topic. Fortunately, with the popularity of social media, social welfare communities and victim support groups facilitate the victims to share their abusive stories and allow others to give advice and help victims. Hence, in order to offer the immediate resources for those needs, the specific messages from the victims need to be alarmed from other messages. In this paper, we regard intention mining as a binary classification problem (abuse or advice) with the usecase of abuse discourse. To address this problem, we extract rich feature sets from the raw corpus, using psycholinguistic clues and textual features by term-class interaction method. Machine learning algorithms are used to predict the accuracy of the classifiers between two different feature sets. Our experimental results with high classification accuracy give a promising solution to understand a big social problem through big social media and its use in serving information needs of various community welfare organizations.