Crowd annotation of cultural heritage collections

Our cultural heritage may not be among today's blockbusters, and as a consequence, we may have much less consumption and description information about them. However, the people who do look at such heritage collections, are usually also quite fanatic about them. While they already interact with the information, they can give us useful insights about the information's content. Furthermore, also looking at many existing successes in the field of citizen science and crowdsourcing, it actually is possible to engage many people in helping describe complex content, as long as we frame the tasks in accessible ways.

In the cultural heritage domain, with regard to museum collections, earlier studies show that Web users can help museum professionals describe collection items (Dijkshoorn et al., 2013; Oosterman et al., 2015). Connected research challenges involve identifying the niche of relevant experts, motivating them to contribute to the annotation of artworks, evaluating the quality of the annotations and annotators using trust algorithms, and presenting all these aspects in a suitable interface. 

In TROMPA, this concept will be taken to the music domain. Novel methods will be devised to integrate multiple crowds, with different expertise and costs, for a heterogeneous set of crowd tasks. Models and tools will be established to combine crowd tasks with algorithmic tasks in hybrid annotation workflows. 



Dijkshoorn C., et al. (2013), “Personalized Nichesourcing: Acquisition of Qualitative Annotations from Niche Communities”, UMAP extended Proceedings.

Oosterman, J., Yang, J., Bozzon, A., Aroyo, L. and Houben, G.J. “On the impact of knowledge extraction and aggregation on crowdsourced annotation of visual artworks”, Computer Networks, 90, pp. 133-149, 2015.