English Semantic Recognition Based on Intelligent Algorithm

Na Deng

doi:10.31449/inf.v45i6.3728

Abstract

In the process of translation, semantic barriers have attracted extensive attention from researchers. Taking the translation between Chinese and English as an example, this paper used intelligent algorithms to recognize the semantic role of English, introduced the semantic role labeling, designed a semantic role encoder, integrated the encoder with the transformer model, and tested the translation performance of the system. The experimental results showed that the BLEU-4 score of the combined system was significantly higher than the baseline system and the traditional transformer system. The average BLEU-4 values of the three systems were 35.02, 35.78, and 36.9, respectively, and the score of the combined system was the highest. The specific analysis of several examples also found that the translation results of the combined system were more reliable. The experimental results verify the effectiveness of the combined system in machine translation and the importance of semantic recognition in translation.

Highlights

Machine translation refers to translating one natural language into another through machine tools
On NIST 02, the BLEU-4 scores of the three systems were 37.2, 38.1, and 39.7, respectively, and the BLEU-4 value of the transformer system combined with semantic role recognition was 6.72 % higher than that of the baseline system and 4.20 % higher than that of the transformer system
On NIST 03, the BLEU-4 scores of the three systems were 36.6, 37.2, and 38.4, respectively, and the BLEU-4 value of the transformer system combined with semantic role recognition was 4.92 % higher than that of the baseline system and 3.23 % higher than that of the transformer system

Summary

Introduction

Machine translation refers to translating one natural language into another through machine tools. Lee et al [4] mapped the source character sequence to the target character sequence without any segmentation They adopted the character-level convolution network with maximum pooling on the encoder to improve the speed of model training. Through experiments, they found that the method had higher translation quality. The higher the mastery degree is, the smaller the obstacles to semantic recognition in the process of translation are This study analyzed the method of Chinese-English machine translation and carried out experiments on the designed system to understand the reliability of the system in translation and make some contributions to the better development of machine translation

Semantic role labeling

Transformer model combined with semantic role encoder

Experimental data

Evaluation index

Experimental results

Conclusion