Digital resources and computational techniques have received limited attention in art history, in part due to the visual and material nature of many of the primary sources with which art historians engage. My paper aims to contribute to filling that gap, exploring their value through looking at the titles of works of art. Titles are closely associated with the works they name. As the semiotician Jose Besa Cambrupi has discussed, titles also act meta-linguistically, saying something about those works and contributing to the meanings they are given (Besa Cambrupi 2002). As self-contained textual units, titles are well-suited to the sorts of statistical approaches used by scholars to analyse texts. Looking at titles in aggregate allows me to bring these techniques into art history and to explore ways in which those meta-linguistic functions have been elaborated.
Drawing on the online collections of 37 modern and contemporary art museums in 17 countries I have assembled a database of 61,000 titles covering the decades from the 1900s to the 2010s. To investigate and interpret this material I turn to the statistical techniques of correspondence analysis, topic modelling and parts of speech tagging. Using these techniques individually and in combination, I read the language used in titles in terms of a long-term narrative within modern and contemporary art as artistic interests came and went and were re-inflected, and as epistemic perspectives on the kinds of knowledge that art can or should engender changed.
My work shows how computational techniques can bring new ways of seeing and new kinds of knowledge into art history. I will argue they give art historians a way of working with primary sources that are inherently noisy. There are also lessons for those working in the digital humanities from my work.
Key words: Art History, Topic Modelling, Parts of Speech Tagging
References:
José Bepa Cambrupi, Nouveaux actes semiotiques No 82: Les fonctions du titre, (Limoges: Presses Universitaires de Limoges, 2002).