ChineseSimilarityDataSets

This dataset contains benchmark similarity results for large-scale Chinese long-text similarity analysis across lexical, semantic, and hybrid paradigms. The corpus consists of 45 Chinese texts spanning modern literature, Ming–Qing dynasty novels, modern essays, and ancient Chinese essays.For each text, three length variants are provided: long (100%), medium (50%), and short (20%), forming a total of nine experimental subsets. Ground-truth relations are hierarchically defined at four levels: same chapter, same book, same author, and same category.