Accuracy Isn’t All: Testing KuroNet for Kanbun OCR
Bibliographic record
Abstract
This is an accepted article with a DOI pre-assigned that is not yet published.This paper explores the potential for using the open-source OCR software KuroNet for the OCR processing of historical texts written in kanbun. Kanbun, a literary standard for official documents used widely throughout Japan from the classical period into the modern period, remains underserved in digital terms. This is also due to particular challenges specific to this writing standard, which uses Chinese characters and syntax to represent Japanese textual content. KuroNet was developed for a different purpose, namely, to help decipher literature written in cursive Japanese characters; yet preliminary experiments show that KuroNet is also capable of solving some bigger challenges of kanbun OCR. This paper first presents an introduction the challenges unique to OCR for kanbun and the implications of the functionalities specific to KuroNet in that context, followed by a review of the experiments conducted using KuroNet on a variety of kanbun documents. As the results show, an unprecedented error rate below 10% could be achieved for a number of documents. The paper further elaborates on the results, possible reasons for systematic errors and closes with recommendations on further work.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.010 | 0.096 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.005 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".