Scholarlink

<p>The existing machine translation systems are limited in translation quality and speed due to their low efficiency in processing massive data and insufficient semantic capture abilities. The compaction algorithm and self-attention mechanism are combined in this article to achieve efficient processing and semantic understanding of the corpus and improve the performance of the translation system. This article uses Huffman coding to compress the vocabulary and reduce storage requirements, optimizes the semantic capture of corpus through a self-attention mechanism, and enhances the semantic alignment of multiple languages based on the Transformer model. Then generative adversarial networks are utilized to realize cross-lingual knowledge transfer and improve the multilingual processing capabilities of the system. Experimental results show that the translation system designed in this article (Name it CompressTrans) is significantly better than online translation and traditional machine translation systems regarding compression efficiency and system stability, with an average compression rate of 89.74% and stability of 90.04%. Regarding cross-lingual semantic alignment, the average alignment rate of the system reaches 87.5%, and the BLEU score is 0.96, showing high multilingual consistency. The accuracy and fluency scores of the translation system are 9.2 and 9.1 points, respectively, (out of 10 points), which are significantly better than other systems. In terms of response delay, the system in this article performs better than the control groups and can meet the needs of real-time translation. The results show that the designed system has significant advantages in data compression, semantic alignment, translation quality, and response speed, indicating its wide application potential in multilingual translation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Corpus machine translation system based on compaction algorithm and self-attention mechanism model

  • 期刊论文

  • 发表日期：

    2025年6月6日

查看全文
  • Zijing Li,
  • Xianrui Yan,
  • Wenzhe Yang,
  • Shuhua Pan

摘要

The existing machine translation systems are limited in translation quality and speed due to their low efficiency in processing massive data and insufficient semantic capture abilities. The compaction algorithm and self-attention mechanism are combined in this article to achieve efficient processing and semantic understanding of the corpus and improve the performance of the translation system. This article uses Huffman coding to compress the vocabulary and reduce storage requirements, optimizes the semantic capture of corpus through a self-attention mechanism, and enhances the semantic alignment of multiple languages based on the Transformer model. Then generative adversarial networks are utilized to realize cross-lingual knowledge transfer and improve the multilingual processing capabilities of the system. Experimental results show that the translation system designed in this article (Name it CompressTrans) is significantly better than online translation and traditional machine translation systems regarding compression efficiency and system stability, with an average compression rate of 89.74% and stability of 90.04%. Regarding cross-lingual semantic alignment, the average alignment rate of the system reaches 87.5%, and the BLEU score is 0.96, showing high multilingual consistency. The accuracy and fluency scores of the translation system are 9.2 and 9.1 points, respectively, (out of 10 points), which are significantly better than other systems. In terms of response delay, the system in this article performs better than the control groups and can meet the needs of real-time translation. The results show that the designed system has significant advantages in data compression, semantic alignment, translation quality, and response speed, indicating its wide application potential in multilingual translation.