Recipient organisationCardiff UniversitySource-published name: Cardiff University
Funding£589K
PeriodAug 2025 — Aug 2028
In plain English
AI plain-English summary
Large language models like ChatGPT are reshaping how people work, but only big tech companies with vast computing power can build and run them. This fellowship aims to change that by making advanced language technology accessible to researchers, public bodies, and small organisations who lack those resources. The core problem is that the most powerful NLP models are expensive to train and run, creating a divide between those who can afford them and everyone else. Meanwhile, fields like healthcare and environmental science generate enormous amounts of text data that humans simply cannot process manually. Without accessible tools, valuable insights remain buried. The researcher is developing efficient methods that require less computing power, releasing open-source models and datasets, and studying exactly what these models can and cannot do. If successful, a public health researcher could analyse millions of patient records without a corporate supercomputer, or a local council could process public feedback at minimal cost. This is applied research with a clear social mission: democratising access to language AI so that its benefits spread beyond a handful of wealthy corporations. The work combines practical tool-building with fundamental research into model efficiency and limitations.
View original technical description
The vision of my FLF fellowship was to offer a bridge between AI theory and practice when it comes to language technologies. The last few years during the fellowship have been paramount to the progress of the field. The general public are becoming more familiar with language technology (e.g. through applications such as ChatGPT) and are increasingly aware of what it can offer. To this end, the first years of the FLF are contributing to shaping this landscape, in particular in relation to the development of efficient and practical models. For the next three years, I will focus on the practical and multi-disciplinary components that the initial research has contributed to, while also reinforcing and extending the foundational research behind this progress. Natural Language Processing (NLP) as a discipline is constantly evolving, rapidly so especially in the last few years. This speed brings about potential risks but also opportunities to harness these advances to put and bring them to provide social good. My vision takes a positive view on this technology and the opportunities that it can bring to many important disciplines, including, but not limited to, health and environment. Data, and in particular textual data, is constantly growing in all disciplines, and makes it impossible for humans to process and get insights from them. Here is where NLP comes into play, but nowadays only experts with access to expensive computing resources (usually large corporations) can take advantage of the exponential progress in the field. The vision of my fellowship is aimed at facilitating the access of NLP technology to non-expert users, including (but not exclusively) researchers from other disciplines, industry and public bodies for positive benefit. This can be achieved by researching efficient methods, better understanding the strengths and limitations of these models, and by releasing open data and specialised models. To fulfil this vision, the renewal of the FLF will allow me and my team to continue to develop the research, keep pace with the rapid evolution of the field, allow us to continue the work started and offer the space to explore exciting new collaborations and applications. Crucially, the practical and multi-disciplinary nature of the scheme, coupled with the strong research foundations required to achieve these goals, make the FLF scheme unique in helping achieve these goals.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know