Combating Human Trafficking with Multimodal Deep Models

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1547–1556 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1142 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1547–1556 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1142 Combating Human Trafficking with Deep Multimodal Models Edmund Tong* Language Technologies Institute Carnegie Mellon University edtong@cmu.edu Amir Zadeh* Language Technologies Institute Carnegie Mellon University abagherz@cs.cmu.edu Cara Jones Marinus Analytics, LLC cara@marinusanalytics.com Louis-Philippe Morency Language Technologies Institute Carnegie Mellon University morency@cs.cmu.edu Abstract Human trafficking is a global epidemic af- fecting millions of people across the planet. Sex trafficking, the dominant form of hu- man trafficking, has seen a significant rise mostly due to the abundance of escort web- sites, where human traffickers can openly advertise among at-will escort advertise- ments. In this paper, we take a major step in the automatic detection of advertise- ments suspected to pertain to human traf- ficking. We present a novel dataset called Trafficking-10k, with more than 10,000 ad- vertisements annotated for this task. The dataset contains two sources of informa- tion per advertisement: text and images. For the accurate detection of trafficking ad- vertisements, we designed and trained a deep multimodal model called the Human Trafficking Deep Network (HTDN). 1 Introduction Human trafficking “a crime that shames us all” (UNODC, 2008), has seen a steep rise in the United States since 2012. The number of cases reported rose from 3,279 in 2012 to 7,572 in 2016—more than doubling over the course of five years (Hot- line). Sex trafficking is a form of human trafficking, and is a global epidemic affecting millions of peo- ple each year (McCarthy, 2014). Victims of sex trafficking are subjected to coercion, force, and con- trol, and are not able to ask for help. Put plainly, sex trafficking is modern-day slavery and is one of the top priorities of law enforcement agencies at all levels. A major advertising ground for human traffickers is the World Wide Web. The Internet has brought * Authors contributed equally. traffickers the ability to advertise online and has fostered the growth of numerous adult escort sites. Each day, there are tens of thousands of Internet advertisements posted in the United States and Canada that market commercial sex. Hiding among the noise of at-will adult escort ads are ads posted by sex traffickers. Often long undetected, traffick- ing rings and escort websites form a profit cycle that fuels the increase of both trafficking rings and escort websites. For law enforcement, this presents a significant challenge: how should we identify advertisements that are associated with sex trafficking? Police have limited human and technical resources, and manually sifting through thousands of ads in the hopes of finding something suspicious is a poor use of those resources, even if they know what they are looking for. Leveraging state-of-the-art machine learning approaches in Natural Language Processing and computer vision to detect and re- port advertisements …

Content truncated to 500 words