Unstable Data and the Unusual Case of the Prosody Excerpt in the Digital Library

Stable data is essential for repeatable research, and cultural heritage data is an invaluable resource for computational humanities research, but the fluctuations within this kind of data pose challenges to reproducible, repeatable, and follow-up research. We use the case study of HathiTrust Digital Library content within the Princeton Prosody Archive (PPA), particularly excerpted and augmented content, as a window into the surprising instability of this large-scale data. We analyze the rate of change in PPA excerpted content, all of PPA, and all of HathiTrust resulting from HathiTrust updates. We use this case study to illuminate the degree of change in HathiTrust, as one exemplar of a cultural heritage data aggregator, which we think is not well understood by computational researchers, and to consider the implications for research. This poster was presented at the Computational Humanities Research (CHR) conference at the University of Luxemberg on December 11, 2025. It summarizes the findings of an accompanying short paper, which can be found in the conference proceedings, published in the Anthology of Computers and the Humanities: https://doi.org/10.63744/cTWwVSItf41f.