A Twitter-Based Analysis for Conversations on the Topics Related to Violence Against Women

Background: Twitter has fostered an increase of online activities that lead to a great amount of social interactions and information transfer for critical public health issues, such as violence against women. Due to its large amount of real time and easily accessible data, Twitter has been utilized for quantitative research, such as sports, disease outbreak, NGO advocacy, marketing and psychological well beings. However, no previous studies have investigated messages and conversations on the issue of violence against women.

Purpose: This study aims to explore the tweets, retweets and conversation structures on the topic of violence against women on Twitter.

Methods: This study use data mining and statistical techniques. We use Twitter API to collect data of ~2.5 million tweets from across an 8-year period (2007–2015). Snowball sampling method is used to pull tweets using a breadth-first crawl of users' “following” structure, starting from 3 random initial Twitter users as seeds. Sample of this study is a specific, self-selected population who use Twitter.

Metadata relating to “reply-to” and “retweet” structure is collected for each tweet. We utilize a “hashtag (prefixed by the # symbol)” based approach to classify conversational topic within tweets. A list of relevant hashtags related to the topic of violence against women is manually generated, such as #SpousalAbuse, #DateRape, #wifebeating, #DomesticViolence, etc.

The conversational structure is measured by (1) the number of tweets per thread, (2) the number of users per thread, (3) the thread depth, and (4) the “thread degree”.  We also select some topics as a comparison regarding the conversational structure with the topic of violence against women, such as politics, sports, fashion and kids. These four measures of conversational structure are collected and analyzed.

Results: Results show that the topic of VAW has far higher values on average number of tweets and users per thread, which emphasize increased activity on this topic. However, the average number of retweets on the topic of VAW is significantly smaller than the majority of other topics. These findings reflect a “Whisper! Don't shout!” conversation structure, and mirror findings from past non-online sociology studies that women are reluctant to speak of abuse they had experienced – especially to individuals not close to them.

Conclusion and Implications: We find that there is an increasing activity of the conversations on the topic of VAW but the amount of retweeted messages is far less than the majority of other compared topics. Based on these interesting observations, we are motivated to further investigate social media diffusion dynamics within domestic violence area, and therefore gain a better understanding of society’s view and reaction to these social issues. This study will inform scholars in the field of violence against women as to whether social media could be a potential medium for future violence intervention and help in developing programs and policies.