Yesterday we kicked off our fall 2022 edition of the DH Virtual Discussion Group for ECRs in Belgium. This edition marks our fifth semester of the discussion group, which is jointly organized this year by Prof. Margherita Fantoli (KU Leuven Faculty of Arts), Dr. Sven Lieber (KBR), Prof. Julie M. Birkholz (KBR Digital Research Lab and Ghent CDH), and myself, Dr. Leah Budke (KU Leuven Libraries Artes). Our first session featured a behind-the-scenes presentation by PhD candidate Paavo Van der Eecken from the University of Antwerp. We had a total of 15 attendees during this group meeting, which contributed to a great networking environment and stimulating discussion inspired by Paavo’s work.
Paavo’s presentation, “Viewing Between the Lines: Annotating Sensitive Attributes in Illustrated Children’s Literature,” was based on his PhD project focusing on nineteenth and early twentieth-century children’s literature in Dutch. More specifically, Paavo’s project examines representation in illustrations along the lines of age, race, class, and gender. He shared insight with us into how he developed the annotation strategies for the images in his corpus, which was provided by the DBNL and comprises more than 1000 children’s books.
We learned from Paavo’s presentation that conducting large-scale analyses like his can be greatly aided by the use of digital tools, in this case image annotation tools. For the annotations, Paavo explored the use of a number of different annotation tools, and he stressed the importance of taking time to try different tools. In the end, Paavo settled on the VGG Image Annotator tool.
While Paavo does some of the annotations himself, he is also assisted by other annotators. Annotating and analyzing a corpus of this size with multiple people contributing requires a standardized approach and the VGG Image Annotator tool allows for this. As Paavo explained, he was able to develop a list of labels to populate the image annotator tool, which annotators could then select from to label the images they were working on. If they happened to come across a problematic image–that is to say, one that cannot easily be categorized–then they log it in a shared spreadsheet for follow-up.
The behind-the-scenes talk generated plenty of food for thought. Our attendees this time had questions about the research workflow of the project, the intricacies of analyzing a concept like (historic) representations of race, and the possibility of automating the image tagging (through the use AI).
Our next meeting will take place on Monday 14 November, 15h00-16h30 CET. Join our mailing list to receive the link on the day of the session! If you would like to see the full schedule for the upcoming sessions including abstracts (post will be updated as abstracts are confirmed), please check here.