Development of a Repository for OCR Models and an Automatic Font Recognition tool OCR-D

OCR 模型存储库和自动字体识别工具 OCR-D 的开发

• 批准号: 394448308
• 负责人: Professor Dr. Manuel Burghardt, since 11/2019
• 金额: --
• 依托单位: Abteilung Buchwissenschaft
• 依托单位国家: 德国
• 项目类别: Research data and software (Scientific Library Services and Information Systems)
• 财政年份: 2018
• 资助国家: 德国
• 起止时间: 2017-12-31 至 2019-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/394448308?language=en
• 关键词: Development; Repository; OCR; Models; Automatic

The project addresses the problem of strongly fluctuating recognition rates of OCR for 16th to 18th century historical prints, limiting the full-text digitization of material created by the VD16, VD17, and VD18 programs.Recognition models trained on modern corpora lacking the specifics of historical prints or historic material without thorough bibliographic analysis, retard recognition rates in comparison to the accuracy now routinely achieved for scans of modern prints.The creation of font-specific corpora on the basis of manual tagging is unrealistic, since both non-trivial knowledge of printing history is necessary and the scalability of such an approach would be insufficient. Due to the repetitiveness of the task, such an approach is also very error-prone. The project will enable the humanities to use OCR in a font-specific manner with limited effort. In order to achieve this the project has three main objectives:The development of an online training infrastructure that allows specific models to be trained for these font groups and at the same time for different OCR software.Development of a tool for the automatic recognition of fonts in digitizations of historical prints. In this case, an algorithm for the recognition of fonts in incunabula is first trained using the ground truth found in the Typenrepertorium der Wiegendrucke. In a second step the fonts are grouped according to their similarity in order to get as few groups as possible while maintaining OCR accuracy.Provision of a model repository, in which developed font-specific OCR models are made available to the public.

该项目解决了16至18世纪世纪历史印刷品OCR识别率波动很大的问题，限制了VD16，VD17和VD18程序创建的材料的全文数字化。在缺乏历史印刷品或历史材料细节的现代语料库上训练的识别模型没有经过彻底的书目分析，与现代印刷品扫描通常达到的准确度相比，识别率会降低。在手动标记的基础上创建特定字体语料库是不现实的，这是因为需要打印历史的重要知识，并且这种方法的可扩展性是不够的。由于任务的重复性，这种方法也非常容易出错。该项目将使人文学科能够以有限的努力以特定于字体的方式使用OCR。为了实现这一目标，该项目有三个主要目标：开发一个在线培训基础设施，允许为这些字体组同时为不同的OCR软件培训特定的模型。开发一个工具，用于自动识别历史印刷品数字化中的字体。在这种情况下，首先使用Typenrepertorium der Wiegendrucke中找到的地面真值来训练用于识别incunabula中字体的算法。第二步，根据字体的相似性对字体进行分组，以便在保持OCR准确性的同时获得尽可能少的组。提供一个模型库，其中开发的字体特定OCR模型可供公众使用。