Georgina Willoughby
Pronouns: She/Her
About
My research project
Context-sensitive AI for text simplification and comprehensionThis research project focuses on advancing Artificial Intelligence for text simplification, to prioritise user-centred accessibility and pragmatic fidelity. Digital information is often locked behind complex language. This can have devastating consequences in domains like healthcare, education, and public services. My work explores how AI can bridge between expert language and public understanding, ensuring essential meaning is not lost.
I am developing context-sensitive AI models that are trained to simplify texts while retaining their original communicative intent, nuance, and coherence. This involves drawing on theories from experimental pragmatics, such as Relevance Theory and Grice's maxims, to guide the AI in identifying and appropriately clarifying implicit language, presuppositions, and context. The goal is to design adaptive systems that meet diverse linguistic and cognitive needs, offering a model for how future AI can be co-designed with accessibility as a core principle.
I am a Leverhulme Trust Doctoral Scholar, supported by the Leverhulme Trust's funding for my research.
This research project focuses on advancing Artificial Intelligence for text simplification, to prioritise user-centred accessibility and pragmatic fidelity. Digital information is often locked behind complex language. This can have devastating consequences in domains like healthcare, education, and public services. My work explores how AI can bridge between expert language and public understanding, ensuring essential meaning is not lost.
I am developing context-sensitive AI models that are trained to simplify texts while retaining their original communicative intent, nuance, and coherence. This involves drawing on theories from experimental pragmatics, such as Relevance Theory and Grice's maxims, to guide the AI in identifying and appropriately clarifying implicit language, presuppositions, and context. The goal is to design adaptive systems that meet diverse linguistic and cognitive needs, offering a model for how future AI can be co-designed with accessibility as a core principle.
I am a Leverhulme Trust Doctoral Scholar, supported by the Leverhulme Trust's funding for my research.
Publications
This paper describes the SurreyCTS submission 1 to the BEA 2026 shared task on lexical difficulty prediction, entered in the Open Track. Our approach progressed from base-line multilingual encoders to a hybrid Rem-BERT architecture with extensive feature engineering , combining semantic funnelling, lexical similarity features, attention-derived signals , and language-aware representations. On our internal production validation set, the best single models achieved RMSE 0.8122 (prod-H) and Pearson correlation 0.8968 (prod-G). A weighted ensemble of the five strongest systems, with weights proportional to inverse squared validation RMSE, was submitted as our final entry. On the official shared-task test set, the ensemble achieved RMSE 1.034, 0.945, and 0.861 for Spanish, German, and Chinese respectively, outperforming the open-track base-line in all three settings and placing fifth among open-track teams.