Talks

Recent talks

16 September 2026 — Τεχνητή Νοημοσύνη στην Εκπαίδευση: τα μοντέλα ΤΝ, αντικαταστάτες ή βοηθοί; Talk (in Greek) at the 2nd Model Junior High School of Athens, on critical literacy for AI output, what it changes in assessment, and moving from user to co-designer of the tools. Slides; interactive exercises.
29 August 2026 — Machine Learning for the Humanities. At the 25th International Congress of Byzantine Studies, Vienna, in the roundtable “6.04 RT — AI and Byzantine Studies”. Program.
9 March 2026 — Learning to Align: Addressing Frequency Distribution Shifts. Research seminar, Department of Statistics, Athens University of Economics and Business. Seminar listing.
24 September 2025 — Explainable Foundation Models for Historical and Subjective Data. Keynote at AI-based Methods for the Humanities, BIFOLD, Einstein Center Digital Future, Berlin. Program.
4 July 2025 — Machine Learning for Ancient Languages: Applications to Byzantine Manuscripts. At “Artificial Intelligence: Hopes and Fears”, 44th Thessaloniki Book Festival. Program.
12 June 2025 — Learning to Adapt: Addressing Character Frequency Distribution Shifts in HTR. SCOOP: Source Codes of the Past, Institute for Advanced Study, Princeton. Program.

Session coordination

8 September 2026 — WG2: Document or Handwriting Classification. Co-coordinated the working group’s conclusions session with Maria Konstantinidou and Paraskevi Platanou at the 2nd SCOOP Exchange Meeting, University of Vienna (7–9 September 2026). Program; SCOOP network.

Recorded talks

Improved text toxicity detection by focusing on polarized texts. In other words, an item doesn't have to be black/white nor gray. One group of people may see it as black while another as white.

A substantial number of Greek papyri documents can only be dated tentatively or in approximation as to their chronological placement, due to the lack of decisive evidence. But an error of only 54 years can be achieved, by applying text regression on their transcriptions.

Following the study of ithaca.deepmind.com, this is an overview of machine learning for ancient languages in the context of an invited talk at the University of Vienna (machines that understand) and related to the ML4AL ACL 2024 workshop.