ThingAI/OmniBook

OmniBook数据集围绕通用文化构建，包含完整作品、历史科学文本、哲学与政治思想、诗歌、戏剧和散文等内容。所有文本均源自古登堡计划，未经修改，仅移除了样板页眉/页脚并进行了近相同版本的去重。数据集具有高保真特性：无OCR噪声，文本由志愿者手动校对，保留了原始标点和强调（如维多利亚时代的斜体、有意使用的古旧拼写）。OmniBook is built around general culture and includes complete works, historical scientific texts, philosophy and political thought, poetry, drama, and essays. All texts are unmodified from their original Project Gutenberg sources except for removal of boilerplate headers/footers and deduplication of near-identical editions. It features high fidelity with no OCR noise, manually proofread texts, and preservation of original punctuation and emphasis (e.g., Victorian-era italics, archaic spelling where intentional).