Grazers in Paris: Notes from biblissima+ conference.

Anguelos and Nicolas attended the Biblissima+ conference in Paris on May 21–22 at Campus Condorcet.
We were pleased to receive insightful feedback from many attendees on the posters we presented—one on fully segmented style analysis models by Anguelos Nicolaou, and another on text line segmentation using Mask R-CNN by Nicolas Renet.
While most talks focused on presenting published data and the infrastructure challenges related to organizing and maintaining that data, there were also engaging discussions on the epistemological challenges that AI introduces to the Digital Humanities, as well as the risks of over-reliance on freemium services in DH projects.
Among the many compelling presentations, the retrospective talk by Dominique Stutzmann and Christopher Kermorvant on large-scale, long-term machine learning analysis of paleographic corpora stood out for its relevance to DiDip. Some key takeaways included:
- Store images locally: IIIF services may break or change over time.
- Organize ML outputs as layered data: This allows better structure and interpretability.
- Treat ML outputs as part of a continuous data pipeline: Rather than static results, outputs should be constantly computable and updatable.
We would like to thank the Biblissima+ team—especially Laurie-Anne Tuaire—for making this event great.
OpenEdition suggests that you cite this post as follows:
anguelos (June 12, 2025). Grazers in Paris: Notes from biblissima+ conference. DiDip. Retrieved November 7, 2025 from https://didip.hypotheses.org/3344

