Invited Talk "From Digitization to AI-Readiness" at the DiHuCo Hub Symposium 2026

On August 7, 2026, the DiHuCo Hub Symposium 2026 — “Toward Agentic DH: Diversifying the Humanities–AI Relationship and Expanding Human–AI Collaborative Research” was held at Hitotsubashi Hall in Tokyo. In a session featuring researchers supported by JST BOOST, the next-generation AI talent development programme, Naoya Iwata gave a talk titled “From Digitization to AI-Readiness: The Design and Sustainability of Humanitext as a Text Infrastructure for Western Classics.”

The talk drew a distinction between digitising texts and making them AI-ready. Transcription and TEI/XML markup have long supported reading, searching, and counting. AI-readiness is the next step: structuring data so that an AI can reach an answer with minimal context and return the identifiers of the passages on which that answer rests, allowing a human reader to check it against the source. Iwata then presented Humanitext’s three-layer architecture — a TEI foundation layer, a text reference layer, and a knowledge-graph reasoning layer — and reported on the current state of Antiqua, Reader, and OCR.

The talk was also candid about operations. By August 2026, cumulative queries to Antiqua and Aozora had passed 140,000, at a cost of roughly 2–3 yen per query. Humanitext has reached this scale with a three-person research team working alongside AI agents; maintaining a public research service over time, however, poses a different set of challenges. By sharing the cost structure as well as the technical design, the talk addressed the symposium’s theme of Agentic DH from the perspective of a team putting it into practice.