Deciphering Ancient Chinese Scripts with Convolutional Neural Networks : A Study on Calligraphic Styles in Classical Chinese Texts

Abstract

The primary objective of this study was to employ deep learning algorithms, specifically convolutional neural networks (CNN), to identify and classify calligraph styles in classical Chinese texts. Samples representing four calligraphic styles from multiple classical documents were collected for the experiments. These samples underwent meticulous preprocessing and were segmented accordingly. Subsequently, a custom CNN model was developed and trained for this dataset. Our evaluation primarily utilized accuracy and confusion matrices to assess the model's performance. Preliminary results showcased the model's excellence in distinguishing between different calligrapher styles. The study suggests that CNNs are highly effective for analyzing complex calligraphic elements in Chinese literature. This approach offers valuable applications in digital humanities and cultural preservation, where such technology can aid in the digitization and detailed study of ancient texts. Moreover, it opens avenues for educational and cultural institutions to leverage this model for enhanced understanding and presentation of calligraphic art.