Learning to Decipher Hate Symbols

Proceedings of NAACL-HLT 2019, pages 3006–3015 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 3006 Learning to Decipher Hate Symbols Jing Qian, Mai ElSherief, Elizabeth Belding, William Yang Wang Department of Computer Science University of California, Santa Barbara Santa Barbara, CA 93106 USA {jing qian,mayelsherif,ebelding,william}@cs.ucsb.edu Abstract Existing computational models to understand hate speech typically frame the problem as a simple classification task, bypassing the un- derstanding of hate symbols (e.g., 14 words, kigy) and their secret connotations. In this pa- per, we propose a novel task of deciphering hate symbols. To do this, we leverage the Ur- ban Dictionary and collected a new, symbol- rich Twitter corpus of hate speech. We investi- gate neural network latent context models for deciphering hate symbols. More specifically, we study Sequence-to-Sequence models and show how they are able to crack the ciphers based on context. Furthermore, we propose a novel Variational Decipher and show how it can generalize better to unseen hate symbols in a more challenging testing setting. 1 Introduction The statistics are sobering. The Federal Bureau of Investigation of United States1 reported over 6,000 criminal incidents motivated by bias against race, ethnicity, ancestry, religion, sexual orienta- tion, disability, gender, and gender identity dur- ing 2016. The most recent 2016 report shows an alarming 4.6% increase, compared with 2015 data2. In addition to these reported cases, thou- sands of Internet users, including celebrities, are forced out of social media due to abuse, hate speech, cyberbullying, and online threats. While such social media data is abundantly available, the broad question we are asking is—What can ma- chine learning and natural language processing do to help and prevent online hate speech? The vast quantity of hate speech on social me- dia can be analyzed to study online abuse. In 1https://www.fbi.gov/news/stories/2016-hate-crime- statistics 2https://www.fbi.gov/news/stories/2015-hate-crime- statistics-released Figure 1: An example tweet with hate symbols. recent years, there has been a growing trend of developing computational models of hate speech. However, the majority of the prior studies focus solely on modeling hate speech as a binary or multiclass classification task (Djuric et al., 2015; Waseem and Hovy, 2016; Burnap and Williams, 2016; Wulczyn et al., 2017; Pavlopoulos et al., 2017). While developing new features for hate speech detection certainly has merits, we believe that un- derstanding hate speech requires us to design com- putational models that can decipher hate sym- bols that are commonly used by hate groups. Figure 1 shows an example usage of hate sym- bols in an otherwise seemingly harmless tweet that promotes hate. For example, Aryan War- rior is a longstanding racist prison gang based in the Nevada prison system. WPWW is the acronym for White Pride World Wide. The hate symbols 1488 and 2316 are more implicit. 14 symbolizes the 14 words: “WE MUST SECURE THE EXISTENCE OF OUR PEOPLE AND A FUTURE FOR WHITE CHILDREN”, spoken by members of the Order neo-Nazi movement. is the 8th letter of the alphabet, so 88=HH=HeilH Hitler. Similarly, W is the 23rd …

Content truncated to 500 words