Experimental Data and Results on Multilabel Classification of Ancient Poetry and Wine Culture

Each sample is annotated with one or more of four cultural labels: institutional culture, material culture, spiritual culture, and behavioral culture.

The dataset also provides summarized experimental results, including dataset statistics, pretrained model comparisons, PC-PLM and feature-concatenation results, category-wise F1 scores, K-sensitivity analysis, imbalance-handling comparisons, raw results from five random seeds, mean and standard deviation summaries, and paired t-test results. The labeled samples were selected and organized from the “Geographical Dataset of Wine Culture in Tang and Song Poetry of the Yangtze River Basin.” The unlabeled corpus used for domain-adaptive pretraining was obtained from the GitHub project chinese-poetry-and-prose and is not redistributed here.