Liu, Ziwen; (2024) Semantic Embeddings of Heritage Sites: A Novel Weakly-Supervised Learning Approach to Analyzing Social Media for Cultural Insights. Doctoral thesis (Ph.D), UCL (University College London).

Text Ziwen Liu - 17033499 - PhD Thesis.pdf - Accepted Version Access restricted to UCL open access staff until 1 November 2025. Download (16MB)

Abstract

Motivated by the transformative role of user-generated content on social media in cultural heritage and the fast-advancing few-shot learning abilities of Artificial Intelligence (AI), this research uniquely positions itself at the intersection of machine learning, heritage studies, and social media analysis, aiming to deepen our understanding of UK heritage sites through the semantic perceptions expressed in online user comments and photographs. Methodologically, this research introduces an innovative weakly-supervised learning framework that significantly reduces the need for extensive manual labelling while enhancing interpretability. This approach allows for the generation of high-quality, interpretable embeddings (numeric vectors) for heritage sites from both textual and visual social media data. This paradigm allows for a measurable understanding of heritage sites, facilitating quantitative analysis like clustering and sentiment analysis. It demonstrates how social media can serve as an effective open-ended surveying tool in visitor experience analysis for heritage sites and public engagement in heritage site management. With the methodological innovations, this study has the potential to provide practical benefits to heritage studies through the development of applications like a recommendation system that highlights diversity and advocates for less-visited heritage locations, thereby supporting the advancement of sustainable tourism. It also provides insights into visitor sentiment analysis and user behaviour modelling, which are crucial for the management and sustainable development of heritage sites. This research can not only contribute to the field of heritage studies, but also pave the way for the application of advanced computational techniques in social science research. The innovations and applications developed in this research hold the potential to deepen our understanding of visitor perceptions towards heritage using AI technologies, setting the stage for future advancements in this exciting interdisciplinary field.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item