BullingerDB: A Dataset for Handwritten Text Recognition and Writer Retrieval
This dataset provides a new benchmark for multilingual historical text recognition and temporally-aware writer analysis, addressing the need for large-scale, diverse historical document datasets.
BullingerDB is a large-scale benchmark dataset of 20,898 pages and 499,222 text lines from 796 writers for historical document analysis. TrOCR achieves a character error rate of 9.1% on text recognition, while writer retrieval yields a mean average precision of 78.3%, highlighting challenges from long-term stylistic variation.
We present BullingerDB, a large-scale benchmark dataset for historical document analysis based on the correspondence of Heinrich Bullinger (1504-1575). The corpus comprises 20,898 pages and 499,222 text lines written by 796 writers over six decades, featuring stylistic variation, multilingual content (mostly Latin and Early New High German) as well as meta-information such as writer identity and time. We evaluate BullingerDB on text recognition and writer retrieval. TrOCR, the best performing model, achieves a CER of 9.1%. For writer retrieval, we introduce a temporal nDCG metric to assess time-aware retrieval. While temporally coherent retrieval is achievable, mAP (78.3%) scores indicate challenges due to long-term stylistic variation. With BullingerDB, we aim to establish a new benchmark for multilingual historical text recognition and temporally-aware writer analysis.