We provide weekly summaries of new tool development and release information related to Digital Humanities (DH).

Web Version of NDLOCR-Lite Released

The web browser version “NDLOCR-Lite Web” of the AI-OCR tool “NDLOCR-Lite” released by the National Diet Library has been made public. Users can easily try OCR processing of images and PDFs in their browser, and since processing is completed locally, images and recognized text are not transmitted externally.

Through parallel processing using WebWorkers (up to 8 threads), recognition processing completes in just a few seconds per page, and a paperback book of around 100 pages can be processed in just a few minutes. While operation has been confirmed on Chrome for Android, it appears not to work on iPhone.

After the release, a bug in the reading order estimation algorithm was discovered, with reports of issues where the order of recognition results is incorrect for horizontal text.

This article is automatically generated by AI. There may be omissions or inaccurate descriptions in the content. Sources include X posts, GitHub updates, Current Awareness Portal.