手工标注的手稿页面数据集

该数据集由雅盖隆数字图书馆中105页手稿组成，每页均经过手工标注，标记了诸如印章、段落和装饰性首字母等视觉元素，共分为十个不同的类别。数据集旨在支持文化遗产应用的未来研究，特别是用于训练和测试元数据丰富模型（MEM）的有效性。MEM是一个概念框架，旨在通过结合细调的计算机视觉模型、大型语言模型（LLMs）和结构化知识图谱来丰富数字化收藏的元数据。数据集的创建过程涉及到对手稿页面进行详细的视觉分析，并利用神经网络模型和LLMs来提取和分类视觉元素。这些提取的信息随后被编码到结构化知识图谱中，并与现有的元数据进行整合。该数据集及其关联的MEM框架可用于解决文化遗产数字化中元数据丰富和语义互操作性方面的挑战，从而提高文化遗产研究的可访问性和互操作性。

This dataset comprises 105 pages of manuscripts from the Jagiellonian Digital Library. Each page has been manually annotated with visual elements including seals, textual paragraphs, and decorative initials, and the full dataset is grouped into ten distinct categories. The dataset aims to support future research on cultural heritage applications, particularly for training and evaluating the performance of Metadata Enriched Models (MEM). MEM is a conceptual framework designed to enrich metadata of digital collections by integrating fine-tuned computer vision models, Large Language Models (LLMs), and structured knowledge graphs. The creation of this dataset involves conducting detailed visual analysis of manuscript pages, and utilizing neural network models and LLMs to extract and classify the annotated visual elements. The extracted information is then encoded into structured knowledge graphs and integrated with existing metadata records. This dataset and its associated MEM framework can be used to address the challenges of metadata enrichment and semantic interoperability in cultural heritage digitization, thereby improving the accessibility and interoperability of cultural heritage research.