This paper proposes a model for English-derived place name recognition and translation using a knowledge graph and a phonetic generation algorithm.By integrating multiple algorithms, the model enhances both recognition and translation accuracy.Experimental results show an AUC of 0.892.With 100 training iterations, the recognition error rate is 1.3%, the translation error rate is 0.8%, and the BLEU score reaches 67.3%, demonstrating strong performance.Practical analysis indicates the model has the lowest time consumption, minimal memory usage, superior classification performance, and over 95% fluency and consistency.The innovation of the research lies in the construction of a bidirectional dynamic interaction fusion mechanism through a knowledge graph, an LSTM algorithm, and a bidirectional matching maximum algorithm, targeting the semantic specificity of English-derived place names.This breaks the traditional one-way static fusion and achieves precise scene-based collaboration and closed-loop optimisation of semantics and phonetics.
Defeng Ma (Thu,) studied this question.