As one of the most typical and influential main cultures of Chinese civilisation, the Yangtze River culture has played a leading role in the rejuvenation of national culture. Knowledge Graphs (KGs) and other technologies for semantic retrieval can be employed to enhance the efficiency of querying Yangtze River cultural data, thereby supporting the protection and inheritance of Yangtze River culture. Therefore, this paper designs a multimodal data-intelligent retrieval model based on knowledge graphs (KGs) and deep learning (DL). Build a knowledge graph (KG) for multi-modal data to gain a deeper understanding of the data and perform efficient information retrieval and knowledge discovery. In response to the problems of complex Chinese character shapes and semantic information in the narrative of the Yangtze River culture, this paper takes the Chinese pre-trained language model ChineseBERT as the semantic embedding layer of the text, integrates glyph and pinyin information, and improves the performance of traditional semantic parsing methods in the subtasks of entity mention recognition and relationship prediction; at the same time, a Convolutional Neural Network (CNN) is used to extract features from image data. According to the above simulation experiments, the model proposed in this paper has good intelligent retrieval accuracy.
