On May 25th, 2024, the First International Ancient Chinese Sentence Segmentation and Punctuation Bakeoff (EvaHan 2024 ) was held in Torino, Italy, which was hosted by CAAI, undertaken by the School of Fine Arts, Nanjing Normal University, the College of Information Management, Nanjing Agricultural University and the School of Economics and Management, Nanjing University of Science and Technology and co-sponsored by CAAI Professional Committee of Language Intelligence, was successfully held in Torino, Italy. It was held online and offline in LT4HALA, the sub-conference of LREC-CoLing2024 (https://lrec-coling-2024.org/) .
Chinese classics carrying a long history and profound cultural deposit are the treasure of world civilization. But it is common that those classics were often engraved without segmentation. It takes a lot of manual or material efforts made by experts or scholars through segmenting or adding punctuation marks. Therefore, the work of segmentation and punctuation for classics plays a crucial role in fostering their creative transformation and innovative development, as well as in the preservation and communication of Chinese culture.
The team from Midu Technology Co., Ltd submitted a system that gained the best overall grade in this evaluation, with F-scores reaching 88.47% and 75.29% respectively for sentence segmentation and punctuation in the closed test. It adopts the strategy combined with example enhancement and decoding optimization, significantly improving the understanding and resolution capabilities of large models in the tasks of sentence segmentation and punctuation for ancient Chinese. This evaluation genuinely advanced the standards of sentence segmentation and punctuation in ancient Chinese, facilitating technological exchange between research institutions and participating teams. In the future, more international evaluations dedicated to ancient Chinese will be organized, furthering the protection, inheritance and innovation of classics.
