This article is automatically generated by AI. The content may contain omissions or inaccurate descriptions. Information sources include X posts, GitHub updates, Current Awareness, and Portal.

We compile weekly information on the development and release of new tools related to Digital Humanities (DH).

Kuzushiji Recognition AI "Soramaru Kuzushiji Recognition" Released

Yuta Hashimoto (National Museum of Japanese History) has released "Soramaru Kuzushiji Recognition," a single-character recognition AI for kuzushiji (historical Japanese cursive script). The system was trained on 860,000 characters published in the Kaggle Kuzushiji Recognition Competition dataset and the Historiographical Institute Kuzushiji dataset, enabling recognition of 3,673 character types. The tool features a specification where the character's facial expression and dialogue change based on the confidence level of the recognition results.

Additionally, this tool has already been integrated into the transcription editor of "Minna de Honkoku" (Crowdsourced Transcription), where it can be utilized as a transcription support feature. The recognition model is published on Hugging Face under CC BY-SA 4.0 license and is available in ONNX format, making it possible to integrate into web applications.