raw_source/: source files used by the active document-level and ABSA pipelines
rating_splits/: active rating-model splits and the shared 1000-sample benchmark
rating_train_augmented.csv: train split used directly by best_cnn_cnn_text_core and best_phobert_phobert_rating_core
rating_val.csv: validation split paired with the active rating train split
absa/: active ABSA balanced splits
tests/: all held-out evaluation files
tests/rating/rating_test.csv: test… See the full description on the dataset page:
https://huggingface.co/datasets/DucAnh204/vietnamese-skincare-reviews.
