Heri-Graphs: A Dataset Creation Framework for Multi-Modal Machine Learning on Graphs of Heritage Values and Attributes with Social Media

Researchers developed Heri-Graphs, a novel framework that constructs multi-modal datasets from social media to analyze heritage values and attributes using graph-based machine learning. By integrating images, texts, geo-locations, timestamps, and social network structures from Flickr posts, the study models visual contents and textual semantics as node features while capturing social relationships and spatio-temporal contexts as links in Multi-Graphs. Tested across three UNESCO World Heritage cities—Amsterdam, Suzhou, and Venice—the framework demonstrated high consistency for semi-supervised learning tasks, offering a scalable approach to map heritage values and attributes. This methodology bridges technical relevance in machine learning with societal interests in urban and heritage studies, paving the way for inclusive heritage management practices globally.

Bai, N., Nourian, P., Luo, R., & Pereira Roders, A. (2022). Heri-Graphs: A Dataset Creation Framework for Multi-Modal Machine Learning on Graphs of Heritage Values and Attributes with Social Media. ISPRS International Journal of Geo-Information, 11(9), 469. https://doi.org/10.3390/ijgi11090469