One of the biggest lessons from building Textura is that historical handwriting isn't "one problem."
A 20th-century personal letter, an official government document, and a Byzantine manuscript all follow different palaeographic conventions.
The latest update introduces category-specific transcription context, allowing users to select the type of document before transcription:
- Personal letters & diaries
- Official document (1821–1945)
- Non-Greek documents
This provides the AI with more appropriate historical context before generating a transcript.
Coming soon:
- Modern handwriting
- Ecclesiastical manuscripts
- Byzantine manuscripts
- Early Modern Greek
Next step: improving transcription accuracy through specialised palaeographic knowledge bases, benchmark datasets, and human correction loops.
Log in or sign up for Devpost to join the conversation.