A Knowledge Graph-Based Approach to Analyzing the Construction of Urban Cultural Images and Multi-Level Symbolic Flows in Translanguaging Practice

In this paper, on the basis of relevant theories, based on the adversarial training of BERT-PGD-BiLSTM-CRF entity recognition model and relationship extraction technique to complete the entity extraction and relationship extraction, and then use the entity linking method that fuses attribute and semantic features (BERT+CBOW+CLS) to complete the construction of the knowledge graph and the supplementation of the knowledge graph, and the data is imported into the Neo4j Storage and Display.The symbols contained in the above knowledge graph for the city cultural image in translanguaging practice are divided into three hierarchical symbols, and the symbols are analyzed in terms of flow.In terms of single language usage, English has the highest proportion (22.57%), and Chinese has the best proportion (63.19%) in the process of urban cultural image construction, highlighting the dominant position of Chinese in urban cultural image construction.During the twenty-year period from 2004 to 2023, the trend of social behavioral symbols growth is significantly higher than that of material and spiritual symbol layers, which fits well with the current social development trend.