Deep Learning Approaches to Detecting Safeguarding Concerns in Schoolchildren’s Online Conversations

Proceedings of Recent Advances in Natural Language Processing, pages 364–372 Varna, Sep 4–6, 2023 https://doi.org/10.26615/978-954-452-092-2_041 364 Deep Learning Approaches to Detecting Safeguarding Concerns in Schoolchildren’s Online Conversations Emma Franklin* Renato Software Ltd. Nottingham, UK e.franklin@senso.cloud Tharindu Ranasinghe* Aston University Birmingham, UK t.ranasinghe@aston.ac.uk Abstract For school teachers and Designated Safeguard- ing Leads (DSLs), computers and other school- owned communication devices are both indis- pensable and deeply worrisome. For their ed- ucation, children require access to the inter- net, as well as a standard institutional ICT in- frastructure, including e-mail and other forms of online communication technology. Given the sheer volume of data being generated and shared on a daily basis within schools, most teachers and DSLs can no longer monitor the safety and wellbeing of their students without the use of specialist safeguarding software. In this paper, we experiment with the use of state- of-the-art neural network models on the mod- elling of a dataset of almost 9,000 anonymised child-generated chat messages on the Microsoft Teams platform. The dataset was manually an- notated into two binary classes: true positives (real safeguarding concerns) and false positives (false alarms) that a monitoring program would be interested in. These classes were then fur- ther annotated into eight fine-grained classes of safeguarding concerns (or false alarms). For the binary classification, we achieved a macro F1 score of 87.32, while for the fine-grained classification, our models achieved a macro F1 score of 73.56. This first experiment into the use of Deep Learning for detecting safe- guarding concerns represents an important step towards achieving high-accuracy and reliable monitoring information for busy teachers and safeguarding leads. 1 Introduction As our lives become ever more digital, tradition- ally “offline” activities are steadily moving on- line, and child safeguarding is no exception. In simpler times, it might have been enough for a schoolteacher to walk up and down a classroom to WARNING: This paper contains offensive examples. *The two authors contributed equally to this work. cast an eye over their pupils, or for a member of staff to oversee breaks in the playground to ensure that no bullying takes place. These days, however, children are often to be found online: when they aren’t using school computers to do their work, they are reading news and social websites, watch- ing videos, messaging one another, and sharing content. As a result, schools are now reliant on specific safeguarding technology to help monitor the online activities of their pupils. So necessary is this technology that the UK’s statutory guidance for schools and colleges on safe- guarding children, Keeping Children Safe in Edu- cation (KCSIE)1, heavily emphasises the dangers posed by the internet in schools and outlines the obligations of staff to ensure that appropriate web filtering and monitoring systems are in place. As a result, such systems are commonplace and are used in schools and colleges across the UK as well as abroad. KCSIE points to a range of online risks to which schools must be vigilant, ranging from harmful web content (e.g. …

Content truncated to 500 words