Synopsis Bibliography Dataset

Bibliographic summaries are an important manifestation of ancient cataloging's "distinguishing academic chapters and examining mirror sources", and have valuable research value. However, existing bibliographic summary resources are scattered and of varying quality, which limits the development of summary research applications. This study follows the four principles of comprehensiveness, authority, copyright protection, and version selection. A total of 47 bibliographic abstracts were selected for digitization and field information extraction, and a summary style bibliographic dataset containing 24 fields including title, volume number, section, version, and responsible party, totaling 59624 records, was constructed. Based on the fine tuned GujiBERT model, automated extraction of 1669117 summary entity information was achieved. The developed dataset provides new data support for in-depth analysis of character characteristics, geographical relationships, and evaluation features in ancient books, thereby promoting the deep mining and sharing of ancient book information.