UNICEF/EVAC DRAGON-S 2020-2024

Summary

DRAGON-S brings together data scientists and linguists to develop new tools for preventing and stoping online grooming of children. Online groomers use language to form a relationship with the children they prey on, to gain and then betray children’s trust. Project DRAGON-S is working to keep children safe from this form of technology-assisted child abuse. This is achieved through the development of two tools:

• Shield: An online training portal for social workers, that will present the latest findings in linguistic research on online grooming
• Spotter: An AI-powered tool that will help police investigators triage the huge amount of digital materials that are siezed during investigation.

The Spotter tool will be based on a novel hybrid model that combines deep learning and the latest linguistic findings on the language and conversational tactics that are used by online groomers. These elements of linguistic knowledge are being integrated into the design of a deep neural network to make it more sensitive to the subtle clues of online grooming.

Partners: Department of Linguistic at Swansea University, TARIAN-ROCU

Funding body: UNICEF/EVAC

Role in the project: Co-Investigator, Lead for Machine Learning and Spotter

Preliminary studies:

• PhD supervision: Integrating corpus linguistic knowledge into deep learning methods, Jay Morgan, 2018-2022
• Cherish DE Escalator fund, worth £10K (PI): Co-developing machine learning and corpus linguistics tools for combating online grooming, 2017
• Cherish DE Escalator fund, worth £10K (Co-I): Profiling online grooming communication, 2019