Modeling Linked Cultural Events: Design and Application

Introduction
This paper discusses the promises and pitfalls of linking historical data on cultural events. Quite a few datasets on historical European music, theatre and film are now publicly available online (Baptist 2017). The ones that contain programming information are, at least to some extent, already event-based. However, they are highly heterogeneous in scale and scope, and they generally do not use the same definitions for, for example, venues, events, or companies. Conceptualizing and embedding cultural events such as concerts or theatrical performances in a linked data framework helps to overcome such issues without forcing an overarching ontology, and it enables researchers to acknowledge the performative and interactive nature of cultural expressions within their (local) societal context (Nijboer and Rasterhoff 2018).
By linking event data internally as well as to external knowledge bases such as DBpedia and Wikidata by means of shared vocabularies, researchers are invited to systematically analyse cultural life cross-sectorally (i.e. theatre, music), internationally (European comparisons and connections), and contextually (in relation to local social, economic, political and cultural features) (cf.
EPAD: European Performing Arts Dataverse). In this paper we discuss the conceptual and practical requirements for such a linked-data approach on the basis of a series of research projects on historical cinema, musical, and theatrical events in the research program
Creative Amsterdam: An E-Humanities Perspective (CREATE).

Cultural events
Events play a key role in historical scholarship, and have gained even more urgency with the increasing importance of digital resources in humanities research. Many projects on historical events, however, employ them as devices to structure data collections and do not explicitly aim to develop analytical frameworks in relation to event data collection and data modeling (De Boer et al. 2015; Van Hage et al. 2011; Shaw 2013. An exception can be found in a statistical method known as event history analysis, which treats events as dependent variables, seeking to statistically describe, explain, or predict their occurrence (Allison 2004). Most research on (urban) arts and culture, however, does not try to statistically identify variables that predict or explain an event, for example the staging of the opera Norma or the screening of the movie Casablanca. Rather, historians may seek to identify (series of) events that have contributed to, for example, the canonical status of specific expressions or genres, to the shaping of local and international cultural taste cultures, or to the emergence of some places as particularly creative and cultural.
We therefore emphasize that (networks and series of) events should also be considered as independent variables that can help us identify and disentangle processes of cultural change and continuity. Central in this view is the assumption that 1) events can be seen as units of analysis with structural properties (notably, a time and place) with links to, for example, actors, institutions, other events, and local properties, and 2) that these interlinkages are key to analysing their role in shaping, for instance, local cultural or social life (Tilly 2002). Turning individual event datasets into linked data versions would provide instantaneous insight into how much performing arts datasets overlaps, ontologically, with any of the others. This provides a roadmap for integrating these still scattered data and studying them in conjunction. A systematic analysis of cultural events therefore requires a data structure which allows for querying connections.
Linking cultural event data
A first analysis of performing arts datasets demonstrated that normalizing even the most basic data across datasets is tricky and that trying to completely harmonize and link all the relevant datasets is futile (Baptist 2017). Fortunately, the structure of linked data provides a way to transparently query heterogeneous data, without enforcing an overarching ontology. Breaking events down into variables such as ‘people’, ‘venues’, ‘place’, and ‘time’, for instance, circumvents the issue of formally defining a ‘performance’. Linked data also allows researchers to test various different link-ups of two data sets so they can evaluate the results when they adjust their queries. In the case of cinemas, for example, one of the problems is that the typology of cinemas differs across countries and periods. In the Netherlands cinemas are divided into types ‘A’ and ‘B’ according to frequency of screenings; in Flanders the cinemas are classified according to how soon they tend to new films after their premiere. If the data was put into a relational database it would be necessary to ‘reconstruct’ either of the classifications for the other dataset. But linked data, because of its model of loose connections, allows querying both datasets, defining a classification only during the query.
For the datasets on cultural events such as historical musical and theatrical performances we build on a rigorous relational data model by Karel Dibbets et al. for the
Cinema Context database (Van Vliet et al. 2009). All movies (often circulating under various titles), persons and companies in in this dataset have been identified and aligned to a master record, and where possible linked to the well known and well maintained Internet Movie Database (IMDb). We develop this approach for other datasets and by linking data on cultural events to other datasets and to other knowledge bases using shared vocabularies such as

schema.org
and
Vocabulary of a Friend (VOAF). In this paper we illustrate research potential, but also practical issues by discussing a recent project on the establishment of movie theatres in the city of Amsterdam in the early twentieth century. By linking data on the history of cinema and movie-going to local contextual data (e.g. census data, municipal election data), we assess how linked data might be used to analyse how specific local historical characteristics shaped form and function of urban cultural life.

Bibliography

Allison, P. (2004). Event History Analysis. In Hardy, M. and Bryman, A. (eds.),
Handbook of Data Analysis. Sage Research Methods, pp. 369-385

Baptist, V. (2017). Mapping European Performing Arts Databases. Presentation at the symposium
European Performing Arts Dataverse, 9 November 2017, Amsterdam.

http://www.create.humanities.uva.nl/epad

Cinema Context.

http://www.cinemacontext.nl

De Boer., V., Oomen, J., Inel, O., Aroyo, L., Van Staveren, E., Helmich, W., De Beurs, D. (2015). DIVE into the event-based browsing of linked historical media.
Journal of Web Semantics, 35(3), 152-158

European Performing Arts Dataverse (EPAD).

http://www.create.humanities.uva.nl/epad

Nijboer, H. and Rasterhoff, C. (2018). Linked cultural events: Digitizing past events and its implications for analyzing and theorizing the creative city. In Münster, S., Friedrichs, K., Niebling, F. and Seidel-Grzesińska, A. (eds.),
Digital Research and Education in Architectural Heritage. 5th Conference DECH 2017 and First workshop UHDL 2017, Dresden, Germany, 30-31 March 2017, Springer CCIS series, pp. 22-33

Tilly, C. (2002). Event Catalogs as theories.
Sociological Theory 20(2), 248-254

Shaw, R. (2013). A Semantic Tool for Historical Events. In
Proceedings of the The 1st Workshop on EVENTS: Definition, Detection, Coreference, and Representation. Atlanta, Georgia, 14 June 2013, pp. 38–46

Van Hage, W.R., Malaisé, V., Segers, R., Hollink, L., Schreiber, G. (2011). Design and use of the Simple Event Model (SEM).
Web Semantics, 9(2), 128-136

Van Vliet
, H., Dibbets., K., Gras, H.
(2009). Culture in Context: Contextualization of Cultural Events. In Ross, M., Grauer, M., Freisleben, B. (eds.),
Digital Tools in Media Studies: analysis and research
.Transcript Verlag:
Bielefeld, pp. 27-42