ディープラーニングを用いた歴史的手書き文献の自動翻刻 : コーパス開発の効率化に向けて

Material type: 図書

Author: 宮川, 創

Publisher: 関西大学アジア・オープン・リサーチセンター

Publication date: 2022-03-31

Material Format: Digital

Capacity, size, etc.: -

NDC: -

View All

Notes on use

Note (General)：: This paper discusses the differences and suitable uses of three handwritten text recognition (HTR) programs developed in Europe: Transkribus, eScripto...

Search by Bookstore

Holdings of Libraries in Japan

This page shows libraries in Japan other than the National Diet Library that hold the material.

List of Cooperating Institutions and Databases

Please contact your local library for information on how to use materials or whether it is possible to request materials from the holding libraries.

other

Kansai University Repository
Digital
You can check the holdings of institutions and databases with which 学術機関リポジトリデータベース（IRDB）（機関リポジトリ） is linked at the site of 学術機関リポジトリデータベース（IRDB）（機関リポジトリ）.
Kansai University Repository

Search by Bookstore

Bibliographic Record

You can check the details of this material, its authority (keywords that refer to materials on the same subject, author's name, etc.), etc.

Digital

Material Type: 図書
Title: ディープラーニングを用いた歴史的手書き文献の自動翻刻 : コーパス開発の効率化に向けて
Author/Editor: 宮川, 創
Author Heading: 宮川, 創
Publication, Distribution, etc.: 関西大学アジア・オープン・リサーチセンター
Publication Date: 2022-03-31
Publication Date (W3CDTF): 2022-03-31
Alternative Title: Handwritten Text Recognition for Historical Documents through Deep Learning : Towards More Efficient Corpus Development Process
Periodical title: KU-ORCASが開くデジタル化時代の東アジア文化研究 : オープン・プラットフォームで浮かび上がる、新たな東アジアの姿
Pages: 323-336
Text Language Code: jpn
Subject Heading: HTR(手書きテキスト認識)
OCR(光学文字認識)
自動翻刻
コーパス開発
ディープラーニング(深層学習)
関西大学
Kansai University
Target Audience: 一般
Note (General): This paper discusses the differences and suitable uses of three handwritten text recognition (HTR) programs developed in Europe: Transkribus, eScriptorium/Kraken, and OCR4all. It commences with an overview of deep learning, HTR, and OCR (optical character recognition) before progressing to review the three programs of interest from the perspectives of history, developer, accuracy rate, layout recognition (including writing orientation), user experience, and cost. All three programs use deep-learning machine-learning technologies. They have also all been proven to reach accuracy rates of close to one hundred percent when appropriately trained depending on the quality of the images of handwritten text, training data, and validation data. Second, the user experience is very important; Transkribus has the simplest installation procedure and graphical user interface, while OCR4all and eScriptorium require users to have expert computer skills. Third, in terms of cost, users of Transkribus are required to purchase credits to access the system and use HTR models to recognize a new text, while eScriptorium and OCR4all do not rely on credit purchase. Finally, we conclude this paper with an overview of suitable cases for each program.
DOI: 10.32286/00026614
Format (IMT): application/pdf
Source: KU-0440-20220331-29.pdf
https://kansai-u.repo.nii.ac.jp/record/22599/files/KU-0440-20220331-29.pdf
Access Restrictions: 限定公開
Data Provider (Database): 国立情報学研究所 : 学術機関リポジトリデータベース（IRDB）（機関リポジトリ）
https://irdb.nii.ac.jp
Original Data Provider (Database): 関西大学 : 関西大学学術リポジトリ