Google Research @ Sydney Event
Honoured to have been selected to attend the Google Research @ Sydney Event at the first Google research facility in Australia with my colleague and lab mate.
About Me
Naomi Baes is a doctoral researcher at the intersection of Psychology and Natural Language Processing, specialising in modelling conceptual change in socially salient mental health concepts using language models and historical corpora. Supported by the competitive Australian Government Research Training Program Scholarship (Fee Offset and Stipend), she works under the supervision of Professor Nick Haslam and Dr Ekaterina Vylomova. Together, they developed SIBling, a multidimensional framework for evaluating semantic change across three core dimensions—Sentiment, Intensity, and Breadth—and proposed scalable computational methods for its application across the social sciences. Naomi also led the development of LSC-Eval, with her supervisors and collaborators Dr Haim Dubossarsky and Raphaël Merx, a benchmarking framework that uses LLM-generated synthetic corpora to evaluate semantic change detection methods in controlled experiments. Her research applies these tools to corpora from academic psychology, media, and everyday language to model conceptual change and uncover its social and cultural drivers. She has presented at top-tier venues including ACL 2024, ACL 2025 (forthcoming), IC2S2 2025 (forthcoming), and EMNLP (LChange’23), contributes to shared NLP tasks such as SemEval-2025 (multilingual emotion recognition), and serves on the program committee for the 14th Joint Conference on Lexical and Computational Semantics.
Interests
- Computational Linguistics
- Computational Social Science
- Natural Language Processing
- Psychology
- Semantic Change
Education
PhD, Psychology/ Natural Language Processing
University of Melbourne (2023-)
Graduate Diploma in Psychology (Advanced) with Honours
University of Melbourne
Research Program
My research investigates how concepts change meaning over time, with a focus on the mental health domain. With my PhD supervisors, I have developed a novel framework - SIBling - to model and measure lexical semantic change (LSC) along three dimensions that are typically overlooked by existing approaches, which treat LSC as a unitary phenomenon.
- SIBling: A theoretical model that integrates insights from historical linguistics and psychology, reducing six established types of LSC into three core dimensions: Sentiment, Intensity, and Breadth. [Prototype]
- SIB Toolkit: A computational implementation of the framework that quantifies change along these dimensions, as well as related features: salience and thematic content.
- LSC-Eval: An evaluation pipeline designed to validate methods for detecting LSC. It (1) creates LLM-generated synthetic corpora simulating kinds of change, (2) tests detection methods in controlled experiments, and (3) identifies the optimal approach. [Prototype]
- Applications: I apply this framework to trace the historical semantic evolution of mental health-related concepts (e.g., autism, schizophrenia), and to analyse related social and cultural dynamics such as concept creep, pathologisation, and stigmatisation.
This program contributes by:
- Proposing a multidimensional model of conceptual change (SIBling) grounded in psychological and linguistic theory.
- Developing computational tools to operationalise and apply the framework across concepts and domains.
- Establishing a principled evaluation framework (LSC-Eval) for testing LSC methods.
- Demonstrating the value of SIBling through detailed case studies in the mental health domain.
- Laying the groundwork for future extensions across domains (e.g., law, humanities) and languages.
Featured Publications
Relevant Publications
(2025).
A General Framework to Evaluate Methods for Assessing Dimensions of Lexical Semantic Change Using LLM-Generated Synthetic Data.
arXiv.
(2024).
A Multidimensional Framework for Evaluating Lexical Semantic Change with Social Science Applications.
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics.
(2024).
What should we call mental ill health? Historical shifts in the popularity of generic terms.
PLOS Mental Health.
(2024).
The structure and evolution of social psychology: a co-citation network analysis.
The Journal of Social Psychology.
(2024).
A search for commonalities in defining the common good: Using folk theories to unlock shared conceptions.
The British Journal of Social Psychology.
Invited Talks
Recent News
Quick Updates
July 21–24, 2025: Accepted to present at IC2S2'25 Norrköping, the International Conference for Computational Social Science, on my frameworks for modelling, and evaluating methods for assessing, conceptual change: “SIBling” & “LSC-Eval” (Norrköping, Sweden).
New corpus data + scripts now publicly available — see Resources tab.
Committed my recent PhD paper to ACL 2025 - read our preprint here
Serving on the SEM 2025 Program Committee, 14th Joint Conference on Lexical and Computational Semantics (co-located with EMNLP - Suzhou, China).