Social mining for sustainable cities: thematic study of gender-based violence coverage in news articles and domestic violence in relation to COVID-19.

We argue that social computing and its diverse applications can contribute to the attainment of sustainable development goals (SDGs)-specifically to the SDGs concerning gender equality and empowerment of all women and girls, and to make cities and human settlements inclusive. To achieve the above goals for the sustainable growth of societies, it is crucial to study gender-based violence (GBV) in a smart city context, which is a common component of violence across socio-economic groups globally. This paper analyzes the nature of news articles reported in English newspapers of Pakistan, India, and the UK-accumulating 12,693 gender-based violence-related news articles. For the qualitative textual analysis, we employ Latent Dirichlet allocation for topic modeling and propose a Doc2Vec based word-embeddings model to classify gender-based violence-related content, called GBV2Vec. Further, by leveraging GBV2Vec, we also build an online tool that analyzes the sensitivity of Gender-based violence-related content from the textual data. We run a case study on GBV concerning COVID-19 by feeding the data collected through Google News API. Finally, we show different news reporting trends and the nature of the gender-based violence committed during the testing times of COVID-19. The approach and the toolkit that this paper proposes will be of great value to decision-makers and human rights activists, given the prompt and coordinated performance against gender-based violence in smart city context-and can contribute to the achievement of SDGs for sustainable growth of human societies.