The word “data” seems like it strikes fear in the hearts of all scholars in the humanities, especially in a field like art history. While it would be easy to say that the future of art history is inevitably digital, I think that also discounts the work that good data analysis pioneered by art historians can do for the field. Earlier in the semester, we learned how data analysis can inspire creative digital humanities projects and digital collections. These efforts ultimately make art, history, and knowledge more accessible to those around the world. Shouldn’t that be the goal of any good scholarship?
I had this in mind as I read our assigned articles this past week discussing exactly how to extract this good, clean data that could make knowledge an scholarship more accessible. It turns out, the answer is not quite as clear! Scholars disagree on exactly how to organize data and institutions abide by a certain standard of sharing their data. We saw these opposing views of data organization confronting each other in Susan Elizabeth Gagliardi and Constantine Petridis’s article, “Mapping Senufo: Making Visible Debatable Information and Situated Knowledge,” and Karl W. Broman and Kara H. Woo’s, “Data Organization in Spreadsheets.”[1]
The main difference between Gagliardi and Petridis’s article and Broman and Woo’s is how they treat ambiguity. Ambiguity and uncertainty is unavoidable in art history and surfaces often with ancient art or non-western art, like the Mapping Senufo project which involved organizing West African art. For Gagliardi and Petridis, they struggled to adequately assign labels to pieces of art that had an unclear artist or origin, owing to the fact that there may be multiple spellings or names for the artist or locations. Broman and Woo suggest that every cell in a spreadsheet should be filled, even if the information is unknown, and to have a standardized spelling or phrase to describe the same piece of data. This is helpful to eliminate inconsistences across spreadsheets, but it may not be truly reflective of the actual data.
In the case of the Mapping Senufo project, assigning a standardized name to an artist or location eliminates the ambiguity and reduces its cultural complexity. This elimination of ambiguity asserts that only one name is correct and any other spellings are wrong, imposing this binary worldview and a misleading authority of certainty. If a date or exact artist’s name is unknown, it would be dishonest to represent an approximation or educated guess as authoritatively true.
Including this ambiguity of the unknown is not conducive to using the data for further analysis, though. In class, we saw how one museum, the Tate in London, combatted this in their publicly available API. The Tate includes a qualitative column next to their quantitative information to explain how and when they simplify their data to format it best for analysis. For example, if a painting was created circa 1840 to 1850, they would include that information in the qualitative column, while a more concrete date, “1840,” is added in the quantitative column. In this way, the museum can be transparent about the information they have while also providing the best approximated information for data analysis.
The key takeaway that I gleaned from this topic of data and data analysis from a humanities perspective is that ambiguity is ok—and necessary—in a field like art history. In an environment where data analysis seems to be an exciting new path towards future art historical scholarship, it may be necessary to artificially “clean” ambiguous data for analysis. As long as the data originator is transparent about ambiguity and uncertainty in their data, I think that it can be responsibly simplified for helpful data analysis.
Text Analysis
Text analysis is another method that can be helpful for humanities research, including art history. In class this week, we reviewed the Google N-Gram Viewer and the Voyant text analysis software. Even simple functions like counting word frequency or word placement can be helpful for pinpointing main ideas, subjects, or terminology that occur frequently across a corpus of documents.
For understanding broad trends across all published works, I think that the Google N-Gram word frequency tracker is especially useful. Like most digital tools used in humanities research, this tool identifies patterns, but it cannot provide an explanation for that discovered pattern. For example, in class we discussed how certain words may simply appear less in the 19th century simply because there were less publications compared to the most recent decade. This tool may be helpful for tracking the popularity of certain ideas or concepts, but it is still just a starting point for deeper analysis.
Voyant seems especially useful for scholars who have several text documents to sift through. Using an example article from a library science course, I found phrases and terms that recurred often throughout the text. It was helpful to see phrases that occurred the most often and where they appeared throughout the text.
For a research paper, I am examining the book plates in Owen Jones’s 1856 book, The Grammar of Ornament. I think that text analysis with Voyant will be appropriate to use for this project. Before using the Voyant application, I would need to ensure that the PDF of Jones’s digitized work has been properly OCR’d so that it recognizes the text. Then, I can use Voyant to see what words and phrases that Jones uses frequently. This can give me a better idea, not only of Jones’s work, but also of the language that his contemporaries may have used in related texts.
To further explore Jones’s The Grammar of Ornament, I am considering creating a more comprehensive searchable collection of the book plates in the text. I began this project by uploading ten book plates into CollectionBuilder’s GitHub. To continue this project, I would need to add to the metadata organized in a Google Sheet, including the standard “tombstone” data along with the file identifier. This data can then be converted into a CSV file to be uploaded into CollectionBuilder. This project would not be possible without accessible, data analysis tools.
Incorporating digital tools into the field of art history does not reduce the role of the scholar. In fact, they can aid in analysis and increase accessibility to this scholarship. There are also creative aways to maintain art history’s ambiguity or at least methods to maintain transparency. Through clean data and data analysis, art historians can integrate digital tools into their methodology, continuing make their scholarship more accessible and innovative.
[1] Karl W. Broman and Kara H. Woo, “Data Organization in Spreadsheets,” The American Statistician 72, 1 (2018): 2-10.; Susan Elizabeth Gagliardi and Constantine Petridis, “Mapping Senufo: Making Visible Debatable Information and Situated Knowledge,” in Perspectives on Data, ed. Emily Lew Fry and Erin Canning (Art Institute of Chicago, 2022).
Leave a Reply