Twitter Mining for Detecting Home Violence

Home Violence (HV) is any pattern of conduct intended to obtain or retain power and control over an intimate partner in any relationship. Identifying the HV occurrences will be critical in reducing them. Social Media (SM) is a vital and information-rich tool. HV-related information could well be detected using SM. The research proposed a method to detect HV-related Tweets. More than 20,000 Tweets were retrieved between May 2019 to May 2020 using Twitter API. Data preprocessing and word embedding were done, respectively. Then, to construct the model, the data set was split into training and testing datasets to detect HV-related Tweets; a deep learning model, LSTM, with different word embedding techniques, was used in this research (LSTM+TF-IDF, LSTM+BOW, LSTM+Word2Vec, LSTM+GloVe and LSTM+BERT). With an accuracy of 89.56%, the LSTM+BERT model surpassed the other implemented models.