Overview of EvaHan2025: The First International Evaluation on Ancient Chinese Named Entity Recognition
Bin Li, Bolin Chang, Ruilin Liu, Xue Zhu Zhao, Shen Si, Lihong Liu, Yan Wei Zhu, Zhixing Xu, Weiguang Qu, Dongbo Wang · 2025
Ancient Chinese books have great values in history and cultural studies.Named entities like person, location, time are crucial elements, thus automatic Named Entity Recognition (NER) is considered a basic task in ancient Chinese text processing.This paper introduces EvaHan2025, the first international ancient Chinese Named Entity Recognition bake-off.The evaluation introduces a rigorous benchmark for assessing NER performance across historical and medical texts, covering 12 named entity types.A total of 13 teams participated in the competition, submitting 77 system runs.In the closed modality, where participants were restricted to using only the training data, the highest F1 scores were 85.04% on TestA and 90.28% on TestB, both derived from historical texts, compared to 84.49% on medical texts (TestC).The results indicate that text genre significantly impacts model performance, with historical texts generally yielding higher scores.Additionally, the intrinsic characteristics of named entities also influence recognition performance.It remains challenging to further enhance model recognition performance and to effectively integrate entities from different annotation schemes into a unified system.