Senso at RANLP 2023: Spotlight on Safeguarding
Since 1989, Recent Advances in Natural Language Processing (RANLP) has been one of the foremost international conferences on all things AI, computational linguistics, and natural language processing. This year, we were proud to attend as a sponsor of the conference, as well as presenting an original research paper on using Deep Learning for safeguarding. Read on to find out more!
Natural Language Processing (NLP) might not be at the forefront of everyone’s minds in safeguarding and edtech, but it should be. If you use a smartphone, an AI assistant, a chatbot such as ChatGPT, a spell-checker, or a search engine, you’re making use of NLP. Anything that involves the processing of natural, i.e. human, language, involves a range of linguistic theories, models and techniques developed by the NLP research community. As part of our commitment to linguistics at Senso and our vision of developing the most linguistically accurate safeguarding tool in the industry, attending and sponsoring RANLP 2023 was a no-brainer for us.
RANLP: The Place to Be
RANLP (pronounced ran-ell-pee) brings together the world’s best researchers in NLP and computational linguistics every other year, with keynote talks from internationally renowned experts in both academia and industry.
This time, it was hosted in Varna, Bulgaria, on the stunning Black Sea coast. Despite the sunbathing weather and a string of very tempting restaurants and beach bars in Varna, attendance at the conference sessions was high and the atmosphere was buzzing. It was a packed program spread over three days, with cutting-edge keynotes in the morning and afternoon, four parallel sessions running at once, and fantastic social events every night. We learned that there were more than 160 papers, demos and posters at the conference, with representation from 32 countries! Even the coffee breaks were brimming with brilliant ideas, and we were thrilled to be able to discuss our safeguarding technology with some of the world’s best minds in computer science and NLP. It truly was a privilege to be there!
“Sponsoring the RANLP conference on linguistics and deep learning was a journey of the unknown, and in the end, it left me exhilarated. I'm eagerly looking forward to shaping the future of safeguarding.”
Michael Payne | Director of Operations | Renato Software Ltd.
Putting Safeguarding on the Agenda
As well as attending as a sponsor, we were also proud to present an original research paper on using Deep Learning for child safeguarding in schools. You can find it on page 364 of the conference proceedings, here. In our paper, we demonstrate how machine learning, specifically state-of-the-art deep learning techniques, can be used to successfully identify true safeguarding violations in children’s online messaging. To our knowledge, this is the first ever paper of its kind, and we received a lot of interest and positive feedback from the attendees. Most importantly, we brought the topic of school safeguarding to RANLP and opened up fruitful discussions around what this means for educational technology, really putting this problem on the NLP agenda. Watch this space for future developments on this at Senso!
The paper was a collaboration with Dr Tharindu Ranasinghe, a machine-learning expert at the University of Aston, who was also a keynote speaker at the conference as well as co-author of a number of international submissions. It’s no surprise that Tharindu won the award for “most prolific author” at RANLP!
Bringing Academia and Industry Together
The final session of the three-day conference – saving the best until last – was our sponsor presentation to the conference delegates on what we do at Senso, and why. We described the power of Senso Safeguard Cloud, the children and schools we have helped to protect, and where we’re heading in our linguistic research. The main feedback we received was that we’re doing something truly important and impactful, something that all academics want to achieve with their research.
Many also emphasized the scope for innovative cross-over between the two worlds in making safeguarding technology more intelligent than ever, while also providing real-life use cases for the latest NLP research. We couldn’t agree more! We also received some really insightful suggestions on how we might enhance our technology, including harnessing the power of machine learning and large language models. We were incredibly fortunate to be able to benefit from this kind of expertise.
“The opportunity to fully understand the potential of Senso's safeguarding capabilities was really exciting and so much fun. It’s a huge motivator to keep pushing towards our goals as a company"
Kevin Sheppard | Director of Sales | Renato Software Ltd.
What Did We Learn?
RANLP 2023 was chock-full of fascinating, thought-provoking and cutting-edge presentations. We learned about the latest computer vision research at Google, current challenges in hate speech and misinformation detection, how to undertake responsible machine learning, and the limitations of large language models (LLMs) like ChatGPT. We were struck by the number of presentations that touched on ethical, social, and educational challenges in this new age of LLMs and neural networks, and we were especially inspired by Prof Ed Hovy’s talk on old vs. new NLP. The possibilities of NLP and AI for improving safeguarding – and many other important jobs – is practically limitless, and we left the conference with our heads buzzing full of ideas.
Finally, we learned that blue raspberry lollies go down very well at a conference. Very well indeed.