Automatically Processing Tweets from Gang-Involved Youth: Towards Detecting Loss and Aggression

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2196–2206, Osaka, Japan, December 11-17 2016. Automatically Processing Tweets from Gang-Involved Youth: Towards Detecting Loss and Aggression Terra Blevins Department of Computer Science Columbia University New York, NY, USA tlb2145@columbia.edu Robert Kwiatkowski Department of Computer Science Columbia University New York, NY, USA rjk2147@columbia.edu Jamie Macbeth Department of Electrical and Computer Systems Engineering Fairfield University Fairfield, CT, USA jmacbeth@fairfield.edu Kathleen McKeown Department of Computer Science Columbia University New York, NY, USA kathy@cs.columbia.edu Desmond Patton School of Social Work Columbia University New York, NY, USA dp2787@columbia.edu Owen Rambow Center for Computational Learning Systems Columbia University New York, NY, USA rambow@ccls.columbia.edu Abstract Violence is a serious problems for cities like Chicago and has been exacerbated by the use of so- cial media by gang-involved youths for taunting rival gangs. We present a corpus of tweets from a young and powerful female gang member and her communicators, which we have annotated with discourse intention, using a deep read to understand how and what triggered conversations to escalate into aggression. We use this corpus to develop a part-of-speech tagger and phrase ta- ble for the variant of English that is used, as well as a classifier for identifying tweets that express grieving and aggression. 1 Introduction The USA has the highest rate of firearm related deaths compared to other industrialized countries. Vio- lence is particularly prevalent in cities like Chicago, which has seen a 40% increase in firearm violence in 2015; someone is shot every 2-3 hours in the city. The Chicago Police Department claims that gang violence is exacerbated by taunting between gang members on social media. Recent studies have shown that the new “digital street” is likely to have consequences for one’s lived experiences (Moule et al., 2013; Patton et al., 2013; Pyrooz et al., 2015). Gangs that are highly organized have an increased likelihood of engaging in online behaviors that may include harassing others via the web (Moule et al., 2013). In this paper, we work with a dataset of tweets posted by a young and particularly powerful female Chicago gang member, Gakirah Barnes, and people with whom she communicated. We use the dataset to develop a system that can automatically classify tweets as expressing either loss, grieving the death of friends or family who were shot, or aggression, threatening to harm others often in retribution for a loss. Tweets that don’t fall into either category are classified as other. The ultimate goal of our work is to alert community outreach groups when aggressive tweets are identified so that they can intervene to alleviate a potentially violent situation. We are also interested in enabling interventions when youths are traumatized before grief turns to retribution. Our team includes social workers who labeled the data with discourse tags representing the intention behind the tweet, and computer scientists who developed the classification system in close consultation with the social workers. The language used in the tweets is quite different from Standard …

Content truncated to 500 words