2026-10-14T15:44:22Z https://www.scidb.cn/oai

10.57760/sciencedb.012ky 2026-09-20T10:15:39Z

2026-09-20 A reproducible benchmark and fixed-budget conversion resource for Chinese script selection doi:10.57760/sciencedb.012ky en This dataset provides a reproducible benchmark for fixed-budget Simplified-to-Traditional Chinese character conversion. It combines public Wikipedia, News Commentary, CLUE TNEWS, AFQMC, OCNLI, LCQMC, CSL and THUCNews sources with derived records containing convertible offsets, OpenCC mappings, target rates, structural-risk annotations, predictions, group assignments, source revisions, checksums and regeneration scripts. Restricted raw text is not redistributed; users retrieve it from the original public providers according to the manifests. The resource supports constrained text transformation, robustness, distribution-shift and reproducibility studies. Chinese script conversion; Simplified-to-Traditional conversion; code-mixed text; benchmark; structural risk; reproducibility You Tingrui Zhang Na Xiong Ling Li Jimei PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank