BiblioPage: A Dataset of Scanned Title Pages for Bibliographic Metadata Extraction

2025-03-25

AI Summary

Manual digitization of bibliographic metadata from historical and contemporary archival documents is inefficient due to heterogeneous layouts and the absence of domain-specific benchmark datasets. To address this, we introduce BiblioPage—the first publicly available scanned title-page dataset tailored for historical and archival bibliographic analysis. It comprises 2,000 title pages from Czech library collections, each annotated with 16 structured metadata categories via precise bounding boxes. BiblioPage reflects real-world diversity across time periods and typographic conventions, filling a critical gap in bibliographic metadata extraction research. We provide an end-to-end evaluation framework for fair benchmarking. Experimental results show that object detection (YOLO/DETR) combined with OCR achieves mAP 52 and F1 59, while vision-language models (Llama 3.2-Vision, GPT-4o) attain up to F1 67. The dataset and evaluation code are fully open-sourced to advance document understanding and information extraction research.

Technology Category

• Natural Language Processing: Information Extraction
• Computer Vision: Image and Video Retrieval
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Manual digitization of bibliographic metadata is time consuming and labor intensive, especially for historical and real-world archives with highly variable formatting across documents. Despite advances in machine learning, the absence of dedicated datasets for metadata extraction hinders automation. To address this gap, we introduce BiblioPage, a dataset of scanned title pages annotated with structured bibliographic metadata. The dataset consists of approximately 2,000 monograph title pages collected from 14 Czech libraries, spanning a wide range of publication periods, typographic styles, and layout structures. Each title page is annotated with 16 bibliographic attributes, including title, contributors, and publication metadata, along with precise positional information in the form of bounding boxes. To extract structured information from this dataset, we valuated object detection models such as YOLO and DETR combined with transformer-based OCR, achieving a maximum mAP of 52 and an F1 score of 59. Additionally, we assess the performance of various visual large language models, including LlamA 3.2-Vision and GPT-4o, with the best model reaching an F1 score of 67. BiblioPage serves as a real-world benchmark for bibliographic metadata extraction, contributing to document understanding, document question answering, and document information extraction. Dataset and evaluation scripts are availible at: https://github.com/DCGM/biblio-dataset

Problem

• Manual bibliographic metadata digitization is slow and labor-intensive
• Lack of dedicated datasets hinders metadata extraction automation
• BiblioPage provides annotated title pages for benchmarking extraction models

Innovation

• Scanned title pages dataset for metadata extraction
• Combines object detection models with OCR
• Evaluates visual large language models performance