Researcher vs. Data? A Conference Report on CHR2022

On December 13th, DiDip presented a poster at Computational Humanities Research 2022 (CHR2022). The event was a melting pot for research into humanities data, featuring a wide range of talks and posters from early career scholars to well-established experts. The conference series had its debut as non-remote and to include a pitched poster-session, giving attendees the opportunity to present their ideas and findings in a less conventional way. #CHR2022 [@CompHumResearch] felt like a meeting of equals – both in the 9am conference room and in the late-night karaoke bar.

The occasion provided a wide-ranging platform for scholars and experts to present their research on various topics. We spotted theoretical frameworks ranging from narratology to cultural ecologies, data ranging from ancient literature to social media, and methods ranging from counting frequencies to neural network ensembles.

A common theme throughout the conference was the importance of interdisciplinarity in computational/digital humanities, with many speakers advocating for the use of atypical approaches from ‘outside the field’ to triangulate and better understand the research object. Another key topic were the challenges and pitfalls of working with text as data, with some talks focusing on the limitations of our understanding of text as language.

As keynote speaker Nina Tahmasebi put it, it is important for researchers to never stop questioning how well our models represent (an alleged) reality, and to acknowledge how methodology can influence interpretations. At this point, placing a higher value on human knowledge and intuition in specific domains was mused upon. An emphasis on careful handling of data was echoed by Lauren Fonteyn [@lauren_bliksem]. She claimed that scholars are sometimes caught in a cycle of self-reflection on methods and have yet to find a satisfactory solution and comprehensive answer to the challenges of text as data.

Much of interest to DiDip was the noticeable consensus in CHR2022 that integration of domain-specific expert knowledge is often key in humanities computing. It bolstered our confidence that we can advance the analysis of medieval charters only on the shoulders of diplomatists. Beyond that, the conference included applications from natural language processing and computer vision that dealt with historical data. As for both areas, we were greatly encouraged by the input from other scholars, consolidating our investment into the pipelines we develop.

Overall, the conference made clear the importance of collaboration and interdisciplinarity around the nexus of computational humanities research. It also emphasized the need for researchers to be cautious and reflective when working with text as data, and to be mindful of the limitations of our models and methods. By publishing (self-critical) work on which tools best suit which data, hence sharing our insights, we can spare other scholars the extra work of figuring out what works and what doesn’t, and help to build a more robust and nuanced understanding of the complexities of our data.