SIX undergraduate students from the Southern University of Science and Technology (SUSTech) developed an algorithm which can automatically recognize oracle bone inscriptions, or Jiaguwen, and provide an interpretation, Shenzhen Economic Daily reported.

Oracle bone inscriptions are known as the earliest form of Chinese writing and were ancestors of modern Chinese characters. These inscriptions were found engraved on excavated animal bones and tortoise shells. The earliest oracle bone inscriptions discovered so far are located in Anyang, Henan Province, which date back to over 3,000 years ago.

Academics have been puzzled reading Jiaguwen. In order to address the issue, the six students used their knowledge to develop an algorithm involving technologies such as image and word processing, deep learning and neural network algorithm, under the guidance of teachers from SUSTech’s departments of humanities and social science and computer science.

The student team first built an oracle bone inscription database; with 558 single words and more than 18,000 images included so far. Then, they adopted the ResNet model to complete image recognition. Their next step was to use the pix2pix model to form a GAN featuring U-Net structure as a method to form Jiaguwen to expand samples and create unknown inscriptions. Finally, they used a search algorithm to compare unknown inscriptions with those recorded in the database.

Meanwhile, the team has a plan to develop a WeChat miniprogram to introduce Jiaguwen knowledge to the public and enhance people’s understanding about the ancient Chinese language, said the report.

Team leader Zeng Ming suggested that the combination of humanities and computer technology studies for algorithm development made them feel the charm of interdisciplinary studies.  (Wang Jingli)