embedding_datasets.pkl for Path Signatures Author Anomaly Detection

Precomputed embedding datasets (dataset0-3) for Path Signatures for Author Anomaly Detection in NLP. Contains RoBERTa stream embeddings, UMAP/Random Projection projections, and path signatures for Project Gutenberg corpus. Used to skip compute_embeddings and project_embeddings when running nlp_demo.ipynb. See https://github.com/KarenShark/Path-Signature_Anomaly-Detection本数据集为面向自然语言处理（Natural Language Processing，NLP）中作者异常检测任务的路径签名（Path Signatures）预计算嵌入数据集（dataset0-3）。数据集包含古腾堡计划语料库（Project Gutenberg Corpus）对应的RoBERTa流式嵌入、均匀流形近似与投影（Uniform Manifold Approximation and Projection，UMAP）/随机投影结果，以及路径签名特征。在运行nlp_demo.ipynb时，使用该数据集可跳过compute_embeddings与project_embeddings的计算流程。详细信息请参阅：https://github.com/KarenShark/Path-Signature_Anomaly-Detection