Andrea Nini
Senior Lecturer (Associate Professor) in Linguistics and English Language
What are your research interests?
My main research interest is how to identify a person from the way they use their language. However, I’m not just interested in how to solve this problem computationally but also on the nature of linguistic individuality. How much are we linguistically unique? And why? These questions are relatively unanswered in modern linguistics, despite their value and significance for forensic purposes.
What is the focus of your current research?
At the moment I’m working on the use of simple (rather than large) language models to extract a representation of the uniqueness of the grammar of a person that can then be used for authorship analysis. The advantage of simple language models like n-gram language models is that these are quite explorable by the analyst and many interesting facts about linguistic individuality are actually emerging from these experiments.
What are some projects or breakthroughs you wish to highlight?
The most interesting breakthrough is that these simple language models seem to outperform pre-trained LLMs in these tasks. I’m currently working on a paper with some colleagues on this and we hope to have this up on arXiv soon.
What memberships and awards do you hold/have you held in the past?
I’m a member of the executive committee of the International Association for Forensic and Legal Linguistics, where I also lead the sub-committee on standards for authorship analysis. I’m also a member of the International Cognitive Linguistic Association and of The Philological Society.
What is the biggest challenge in Digital Trust and Security now?
How to regulate the misuse of Generative AI. There is a lot of debate on regulation of AI in order to avoid catastrophic future scenarios but I think more should be done instead on counteracting immediate threats from this technology. I find the possibility of creating an unauthorised digital twin of a person just by using pictures/recordings/videos of this person to be quite scary for trust and security already.
What real world challenges do you see Digital Trust and Security meeting in the next 25 years?
It is very hard to predict what the world will look like in 25 years because of the fast pace in the evolution of these new technologies. If I have to make a relatively safe bet, though, then I think that the same issue of controlling the spread of misinformation that we have today will still be there but on a much larger scale. Finding reliable ways to prove someone’s identity online is going to be key to solve these problems.