# Efficient OCR for Building a Diverse Digital History

Source: [Efficient OCR for Building a Diverse Digital History](<https://www.sota2.com/research/paper/68fa74401e209131cea60891>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-02T07:02:30.494Z

Authors: Jacob Carlson, Tom Bryan, Melissa Dell

Published: 2023-04-05T20:36:04.000Z

## Abstract

Thousands of users consult digital archives daily, but the information they can access is unrepresentative of the diversity of documentary history. The sequence-to-sequence architecture typically used for optical character recognition (OCR) - which jointly learns a vision and language model - is poorly extensible to low-resource document collections, as learning a language-vision model requires extensive labeled sequences and compute. This study models OCR as a character level image retrieval problem, using a contrastively trained vision encoder. Because the model only learns characters' visual features, it is more sample efficient and extensible than existing architectures, enabling accurate OCR in settings where existing solutions fail. Crucially, the model opens new avenues for community engagement in making digital history more representative of documentary history.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Optical Character Recognition on Japanese Horizontal](<https://www.sota2.com/research/sota/optical-character-recognition-on-japanese-horizontal.md>): Optical Character Recognition; Japanese Horizontal; Character Error Rate: 0.6.

- [Optical Character Recognition on Vertical Japanese tables](<https://www.sota2.com/research/sota/optical-character-recognition-on-vertical-japanese-tables.md>): Optical Character Recognition; Vertical Japanese tables; Character Error Rate: 0.7.

- [Optical Character Recognition on Vertical Japanese prose](<https://www.sota2.com/research/sota/optical-character-recognition-on-vertical-japanese-prose.md>): Optical Character Recognition; Vertical Japanese prose; Character Error Rate: 2.7.

- [Optical Character Recognition on Chronicle American LoCCA (eval)](<https://www.sota2.com/research/sota/optical-character-recognition-on-chronicle-american-locca-eval.md>): Optical Character Recognition; Chronicle American LoCCA (eval); Character Error Rate: 1.5.

- [Optical Character Recognition on Ancient Greek](<https://www.sota2.com/research/sota/optical-character-recognition-on-ancient-greek.md>): Optical Character Recognition; Ancient Greek; Character Error Rate: 4.3.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2304.02737>)

[PDF](<https://arxiv.org/pdf/2304.02737.pdf>)