Abstracts
Seminar: Folkloric Transformations: Folklore Classification with Networks and Deep Learning
We present a series of experiments using a large collection of Danish folklore as a starting point. In the first experiment, we show established results from a network reclassification of the corpus. We then move on to a new method for network decomposition to understand the underlying topologies of a complex corpus, revealing latent features in the corpus that align with changes in social and economic organization of late 19th century Denmark. We use this discovery as a motivation to leverage a fine-tuned language model for the study of Danish folklore to explore the overlap between an existing classification scheme and a latent semantic clustering of the documents, which in turn lends support for an investigation of the simultaneous disappearance of witches in Denmark and the rise of the dairy cooperatives. The methods presented are largely language agnostic, and can be applied to corpora from any language, making them widely applicable to problems in the study of Asian cultures, but also making them adoptable for the study of multilingual corpora.
Workshop: BERTopic and multiple classification
In this workshop, we learn how to install and run Bertopic for the analysis of a pre-classified corpus. We work with an existing corpus of classified news articles to see the intersection between existing classifications and the classifications discovered using BERT. We learn a little bit about the parameters for the model, alternate methods of labeling, and how to save our model for reuse.
Seminar: Computational Approaches to Semantic Change in Historical Texts
Semantic change, the evolution of word meanings over time, offers crucial insights into historical, cultural, and linguistic processes, making it a central topic in the Digital Humanities. Language acts as a mirror to societal shifts, reflecting changes in values, norms, and technological advancements. Understanding how word meanings evolve allows us to trace these transformations and gain a deeper understanding of the past.
This seminar explores how computational methods are revolutionizing our ability to analyse semantic change in historical texts, addressing a core challenge in digital humanities. While advanced computational methods enable us to analyse vast datasets and uncover previously inaccessible patterns, few natural language processing algorithms fully account for the dynamic nature of language, particularly semantics, which is critical for humanistic inquiry. As AI systems develop to better understand historical context and language dynamics, human annotation and interpretation remain essential to capture the nuances of language and its cultural context.
In this talk, I will show how computational and human-centred approaches can be effectively combined to examine semantic change and its connections to cultural and technological developments. I will showcase examples of how semantic change can be analysed across temporal, cultural, and textual dimensions. For instance, the study of nineteenth-century British newspapers reveals how industrialization and mechanization significantly influenced the meanings of specific words, providing valuable insight into the intricate relationship between language and technological progress during this transformative period. Furthermore, I will emphasize how the integration of computational models with human annotation enables a more nuanced and complex analysis of language dynamics. This combined approach not only allows us to examine macro-phenomena, such as broad shifts in word usage, but also makes it possible to study semantic change at the level of individual word senses, revealing subtle yet significant patterns that might otherwise be overlooked.
Workshop: Hands-on Historical Semantics: Designing, Implementing, and Interpreting Computational Analyses
This methodologically-focused workshop provides hands-on experience with implementing and validating computational approaches to semantic change (i.e. meaning change) detection in historical corpora. A core focus will be the crucial role of corpus design in shaping research outcomes. Participants will explore the impact of corpus size, composition, and genre representation on the analysis of semantic change. Participants will learn to implement various distributional methods, including collocation analysis and word embeddings. Following best practices for semantic change detection, the workshop will cover crucial methodological considerations in corpus preparation and model development, with specific attention to the challenges posed by historical data. We will addressing critical challenges in historical computational linguistics, such as handling data sparsity and evaluating model performance. Participants will engage in practical exercises designed to highlight the interplay between corpus design choices and the interpretation of semantic change.
Session Details
| Date | Speaker | Theme | Session | Time |
|---|---|---|---|---|
| 5 March, 2025 | Prof. Timothy Tangherlini | Folkloric Transformations | Seminar | 12:30 – 13:30 |
| BERTopic and Multiple Classification | Workshop | 16:30 – 18:30 | ||
| 6 March, 2025 | Dr. Barbara McGillivray | Computational Approaches to Semantic Change in Historical Texts | Seminar | 12:30 – 13:30 |
| Hands-on Historical Semantics | Workshop | 16:30 – 18:30 |
Venue: Arts Tech Lab (Room 4.35), 4/F, Run Run Shaw Tower, Centennial Campus
• Seminar: Up to 40 HKU participants.
• Workshop: Up to 30 HKU participants.
• Registration is based on a First Come, First Served basis. (Deadline: 3 March, 2025)
About the Speakers
Prof. Timothy Tangherlini
Dr. Barbara McGillivray
Professor, School of Information and Scandinavian Section, UC Berkeley
Prof. Tangherlini’s research centers on folklore and informal culture in Scandinavia, with a primary focus on Denmark. Trained as a folklorist and ethnographer, he has extensively studied the circulation of informal storytelling in both agrarian and urban communities, exploring how stories reflect and shape social, economic, and political changes. He has developed computational methods for analyzing topic change and geographic distribution in large folklore corpora, utilizing descriptive statistics, probabilistic modeling, natural language processing, and historical geographic information systems (hGIS).
His work also includes network methods for understanding dynamic systems, such as storytelling networks, story classification, and networks within narratives, from character interactions in Icelandic sagas to actant-relationship networks in witchcraft stories. He has extended these methods to study literary movements, like the Modern Breakthrough, and individual authorships, such as H.C. Andersen.
Additionally, Prof. Tangherlini examines contemporary Danish cinema, focusing on culturally informed systems of interaction and the use of visual narratives to explore social, economic, and political decision-making. His research extends to Korean folklore and culture, as well as the emerging field of Computational Folkloristics. Current projects include tracking conspiracy theories in social media and analyzing dance sequences in K-pop music videos.
Lecturer in Digital Humanities and Cultural Computation, King’s College London
Dr. McGillivray is a digital humanist and computational linguist. Before joining King’s in 2021, she was a Turing Research Fellow at The Alan Turing Institute and the University of Cambridge (2017–2021). Her prior roles include language technologist in the Dictionary division of Oxford University Press and data scientist in the Open Research Group of Springer Nature. She holds a PhD in computational linguistics from the University of Pisa (2010), a Master’s in Mathematics, and a Bachelor’s in Classics from the University of Florence.
As Editor-in-Chief of the Journal of Open Humanities Data and convenor of the MA program in Digital Humanities at King’s, Dr. McGillivray’s work bridges computational methods and humanistic inquiry, particularly in semantic change and historical linguistics.
Registration for this event has been closed. Thank you for your interest.
Prerequisites for Attendees
- Prior experience with digital humanities is not required.
- Programming or coding knowledge is not necessary.
- Workshop participants should bring their own laptop to obtain and retain learning materials.
For inquiries, please contact:
Email: atlask@hku.hk