Combating Human Trafficking Using Structural Information in Online Review Sites

The applicant proposes to develop a prototype tool that will better identify cases of likely victims of human trafficking and those who victimize them using machine learning and social network analysis. The project will result in a significant new capability for law enforcement.

The project specifically focuses on detecting and differentiating child and adult victims. The resulting prototype will allow law enforcement to search and prioritize 1) trafficking cases by victim and 2) trafficking by specific johns. GTARC and the Georgia Tech College of Computing are partnering with the DeKalb County Police Department (DKPD). The DKPD has expressed its willingness to participate in data annotation and prototype development activities including user testing. Data will be collected from online review websites and then annotated for a variety of phenomena including keywords associated with trafficking cases for adults and for children.

Machine learning models will be trained using a ground truth dataset based on online reviews recovered and processed using these keywords. The resulting models will then be trained and optimized to detect and classify online reviews, according to criteria such as trafficking, adult, and child. To enable law enforcement access to the prototype software, GTARC will create a secure public-facing website that will run a copy of the prototype. The web application will require that user’s login with a username and password. GTARC will rely on a snowball style user identification process where existing users of the system may opt to send links to individuals they deem credible as additional users who should have access to the system. The initial set of users will be DKPD.

At the request of the Government, GTARC will make accounts available to any individuals and groups of users the Government recommends. Training materials offered through a short course format will also be made available.

"Note: This project contains a research and/or development component, as defined in applicable law," and complies with Part 200 Uniform Requirements - 2 CFR 200.210(a)(14).

ca/ncf