AI-OCR Product
OCR service incorporating the latest AI and image processing technologies
FROG AI-OCR
Supports multiple languages and can even convert complex Japanese documents from the Meiji to Showa periods into text!
Digitize regional historical documents and other materials into an electronic library and digital archive, and comply with the Reading Barrier-Free Act!
<Comment from the National Diet Library Next-Generation System Development Laboratory>
The Japanese OCR processing program “NDLOCR,” which is the result of this project, was released as open source on April 25, 2022, from the NDL Lab's official GitHub account (https://github.com/ndl-lab). NDLOCR is an OCR system that allows for additional training by preparing training data, and it will be used to create full-text data for materials digitized by the National Diet Library in the future. In addition to the program, the machine learning dataset used in development (created solely from digitized materials whose copyright protection periods have expired) is also scheduled for release shortly. We hope that this will contribute to improving the overall accuracy of Japanese OCR and encourage many interested parties to make use of it.
*Excerpt from the April 28, 2022 release
Translated with DeepL.com (free version)
Our strengths
We are developing optimal AI solutions in collaboration with the Morpho companies and other partners. Working together with engineers and consultants with backgrounds in technology, we promote DX across the entire scope of the process—from the planning stage to the implementation stage.
Our strengths
We are developing optimal AI solutions in collaboration with the Morpho companies and other partners. Working together with engineers and consultants with backgrounds in technology, we promote DX across the entire scope of the process—from the planning stage to the implementation stage.
Books and magazines published after the 1860s can be recognized with an accuracy rate of over 90%, which is higher than commercially available OCR.
In particular, for modern books and magazines published between the Meiji period and the early Showa period, the reading accuracy is approximately twice that of commercially available OCR
(approximately 40% to over 90%).
Go to the FROG-AI-OCR Service Site
