English Translation Stylistic Features and Syntax Translation with Application of Knowledge Mapping

Abstract

Abstract Aiming at the problems of excessive instantiation complexity, poor interpretability and low generalization ability of knowledge reasoning techniques, this paper unites inductive logic programming ILP with HET neural network to construct a hybrid logic rule and neural network knowledge graph reasoning model - HETIL model. The ILP is utilized to quantify the first-order logic rules, and the multi-layer rule space is constructed by arranging and combining the rules through logic symbols. The rules are instantiated and fed into the HET network, and the attention coefficients aggregate the node features to complete the end-to-end training and generate the rule learning model. Finally, the validity of the model is verified by machine translation experiments, and the results show that the accuracy of the HETIL model in syntactic structure types is more than 0.8 overall. The accuracy in terms of phrase structure reaches 0.879. The average BLEU value of the HETIL model can reach 29.24, which is 1.51 BLEU points higher than the benchmark model. Therefore, the effect of English translation by applying a knowledge graph is better than traditional machine translation.