Recently, an artificial intelligence system dubbed the ‘AI that understands oracle bone script best’ has drawn widespread attention. Developed by a Chinese research team, this AI leverages deep learning and a large-scale ancient script database to accurately recognize, translate, and interpret oracle bone inscriptions—China’s earliest systematic writing system dating back over 3,000 years. These inscriptions, carved on animal bones and turtle shells during the late Shang Dynasty, contain invaluable records of ancient society, religion, astronomy, and more. However, their complex glyphs, frequent damage, and lack of contextual clues have long made interpretation extremely challenging. Trained on hundreds of thousands of oracle bone rubbings and existing scholarly research, this AI can not only identify blurred or fragmented characters but also infer meanings based on context and even help reconstruct missing sections. In tests, its recognition accuracy exceeds 90%, significantly outperforming traditional methods. Moreover, it features interactive learning capabilities, allowing continuous model refinement through expert feedback. This breakthrough offers a powerful new tool for paleographic research and opens fresh avenues for exploring the origins of Chinese civilization. In the future, the technology may be extended to other ancient scripts such as bronze inscriptions and bamboo-silk texts, advancing digital humanities to new heights.
近日，一款号称‘最懂甲骨文的AI’的人工智能系统引发广泛关注。该系统由中国科研团队研发，利用深度学习与大规模古文字数据库，能够高精度识别、翻译并解释距今三千多年的甲骨文。甲骨文作为中国最早成体系的文字，记录了商代晚期的社会、宗教、天文等信息，但由于字形复杂、残损严重且缺乏上下文，长期以来解读难度极高。这款AI通过训练数十万片甲骨拓片和已有研究成果，不仅能自动识别模糊或破损的字符，还能根据语境推测其含义，甚至辅助学者重构缺失内容。在测试中，其识别准确率已超过90%，显著高于传统方法。此外，该AI还具备交互式学习能力，可随专家反馈不断优化模型。这一突破不仅为古文字研究提供了强大工具，也为中华文明起源探索开辟了新路径。未来，该技术有望应用于更多古文字体系，如金文、简帛文字等，推动数字人文研究迈向新高度。
原创文章，作者：admin，如若转载，请注明出处：https://avine.cn/3101.html