PhilologyAI

A text is never one reading.

PhilologyAI explores evidence-led, AI-assisted textual criticism for texts that have outlived their first readers.

Keep witnesses, variants, and inference in view. The research direction is a tool that helps scholars examine the evidence—not a substitute for scholarly judgment.

Explore the approach View source repository
Witness comparison Schematic · illustrative only
  • Witness Areading α
  • Witness Breading β
  • Witness Creading α
Illustrative evidence trace Three witness lines meet at a shared evidence pool before a candidate reading is considered.
Variant under review α / β

A visual sketch of the proposed evidence-pool idea. It is not a model output or a transcription of a real manuscript.

Read the variants. Keep the evidence.

PhilologyAI's research scope centers on three enduring questions: what differs between witnesses, what may be missing, and how a reading can be attributed. Each answer must remain connected to the evidence that supports it.

Textual criticism

Compare variant readings while keeping the witnesses and their relationships visible.

Restoration

Explore possible restorations of damaged or incomplete text, with uncertainty made explicit.

Attribution

Examine attribution as an evidence-based question, not a confident label detached from its sources.

Many perspectives. One evidence pool.

The project brief describes a multi-agent approach: specialist perspectives examine the text, while a shared evidence pool preserves the material behind each interpretation for a scholar to assess.

This is the research architecture described in issue #1; the current repository does not contain a verified PhilologyAI implementation or live demonstration.

  1. Gather witnessesBring the textual evidence into view.
  2. Compare readingsSurface agreements and meaningful variation.
  3. Test interpretationsConsider context, restoration, and attribution as questions.
  4. Preserve provenanceKeep each inference connected to its supporting evidence.

Languages, histories, witnesses.

The issue names four areas for the research program. They span literary and documentary traditions whose evidence survives in very different forms.

  • Classical GreekLiterary and documentary texts
  • LatinTexts across manuscript traditions
  • Ancient ChineseHistorical textual traditions
  • InscriptionsTexts carried by material witnesses

Follow the work to its source.

The code repository is the current public access point. No live demo or paper release is identified in the project materials yet.

Open the GitHub repository

Code citation

PhilologyAI. (n.d.). JADEpuffer_PhilologyAi [Source code]. GitHub. https://github.com/philologyai/JADEpuffer_PhilologyAi

Live demoNot published