Computing similarity between items in a digital library of cultural heritage

Abstract

Large amounts of cultural heritage content have now been digitized and are available in digital libraries. However, these are often unstructured and difficult to navigate. Automatic techniques for identifying similar items in these collections could be used to improve navigation since it would allow items that are implicitly connected to be linked together and allow sets of similar items to be clustered. Europeana is a large digital library containing more than 20 million digital objects from a set of cultural heritage providers throughout Europe. The diverse nature of this collection means that the items do not have standard metadata to assist navigation.A range of methods for computing the similarity between pairs of texts are applied to metadata records in Europeana in order to estimate the similarity between items. Various methods for computing similarity have been proposed and can be classified into two main approaches: (1) knowledge-based, which make use of external knowledge sources and (2) corpus-based approaches, which rely on analyzing the frequency distributions of words in documents. Both techniques are evaluated against manual judgements obtained for this study and a multiple-choice test created from manually generated categories in cultural heritage collections. We find that a combination of corpus and knowledge-based approaches provide the best results in both experiments. ACM Reference Format:Aletras, N., Stevenson, M., and Clough, P. 2012. Computing similarity between items in a digital library of cultural heritage. ACM J. Comput. Cult.

Linking and clustering artworks using social tags: Revitalizing crowd‐sourced information on cultural collections

“…Approaches to similarity computing of cultural heritage can be divided into knowledgebased and corpus-based. The former calculates similarities between content by using external resources, while the latter analyzes the frequency distribution of words within the content (Aletras, Stevenson, & Clough, 2012). Since libraries and archives have most of their collections in the form of documents, they are able to adopt a corpus-based approach which compares and analyzes texts.…”

“…To do this, we constructed a term-document matrix and calculated the similarities by the cosine similarity method. Cosine similarity is a widely used approach in the information retrieval vector model; it treats each document (a set of words) as a vector and measures the similarity between the documents by calculating the cosine of the angle between them (Aletras et al, 2012;Kim et al, 2010;Shang, Zhang, Zhou, & Zhang, 2010). We treated artworks as documents and social tags as terms, and built a tags-artwork matrix containing artworks in columns and tags in rows.…”

“…But proving the meaningfulness of the resulting clusters was not easy given the subjectivity in the interpretation of art. In order to address this problem, we carried out an evaluation inspired by earlier studies that compared the results of similarity computing with human judgment of artworks in order to determine the former's validity (Aletras et al, 2012). In this study, we compared our results with categorization of the identical 39 artworks by two curators at GMoMA, which held an exhibition based on social tagging data (Table 3).…”

Evaluation

“…Related or similar items: a component was developed to compute the similarity between pairs of items [3]. This basic functionality was used within a range of features in the PATHS system, such as clustering related items, forming intra-collection links and providing non-personalized recommendations.…”

Why are these similar? Investigating item similarity types in a large digital library

“…To assess annotation quality, we computed the Pearson product-moment correlation of each annotator against the average of the rest of the annotators, as previously (Aletras, Stevenson, & Clough, 2012;Grieser, Baldwin, Bohnert, & Sonenberg, 2011). We then averaged all the correlations.…”