基于人工智能技术的古文字研究

Ancient Chinese Characters Research Based on Artificial Intelligence Technology

人工智能与古文字学交叉研究十分重要,开展这项研究既需要人工收集和标注大量数据,同时也需结合恰当的技术。在数据处理方面,数据集建设过程中尽量丰富了单字数量以及字图总量。数据中的字图包括拓本和摹本,其中拓本多带有斑点噪声,降低噪声有助于提高文字识别的准确率。数据中古文字隶定体的显示也是要重点解决的问题。在文字自动识别方面,利用了深度学习算法开展智能识别,从实验结果看,准确率达到八成以上,这是在大规模识别任务下达到的效果,证明了利用人工智能技术识别古文字形体是可行的。分析错误数据可以发现,数据量与形近字是影响识别准确率的关键因素。除了识别以外,知识图谱技术也很重要,建设古文字知识图谱一方面可以实现对古文字知识体系的多角度展示;另一方面也可计算字形中偏旁及构形的相似度,智能寻找出字形之间的联系。

Interdisciplinary research of AI and palaeography is very important. To carry out this research, it is necessary to collect and label a large amount of data manually and also incorporate some appropriate technologies. In terms of data processing, the number of single words and the total number of ancient characters has been enlarged as much as possible during the construction of data sets. The word image in the data includes rubbings and copies, and the rubbings are mostly with speckle noise. Image Denoising is conducive to improving the accuracy of character recognition. The display of ancient Chinese characters in the data is also a key problem. In the aspect of automatic character recognition, the deep learning algorithm is used to carry out intelligent recognition. From the experimental results, the accuracy rate is more than 80%, achieved under the large-scale recognition task. It proves that the AI is feasible to recognize the ancient character shape. By analyzing the error data, it can be found that the scale of data and the similar characters are the key factors affecting the recognition accuracy. In addition to recognition, knowledge graph is also important. On the one hand, the construction of ancient Chinese character knowledge graph can realize multi-angle display of ancient Chinese character knowledge system;On the other hand, it can also calculate the similarity of the side and configuration in the glyph, and intelligently find the connection between glyphs.