Research on Information Distortion Control in Real-Time News Translation Based on Machine Translation and Semantic Similarity Evaluation

This study aims to explore and propose a method combining machine translation (MT) and semantic similarity evaluation (SSA) technology to effectively control the information distortion in real-time news translation process. For this purpose, this paper has improved the text preprocessing, neural network structure, and attention mechanism from three aspects, explored two technical routes, and proposed the capsNet-BiGRU text semantic similarity analysis model and the RCNN network with central word attention mechanism. The research results show that the capsNet-BiGRU model has superior performance in calculating accuracy, but it requires longer training time and a large number of training samples; while the RCNN network with central word attention mechanism has excellent performance and more efficient calculation. The capsNet-BiGRU model combines capsNet and BiGRU, utilizing the advantages of both to extract multi-dimensional text semantic features. Finally, the text semantic features are used to complete the similarity calculation task. The RCNN network with central word attention mechanism improves the RCNN structure by using BiGRU for semantic feature learning. By using central word attention mechanism and character concatenation, it obtains a semantically enhanced representation of the sentence, and then uses a 1D CNN to fuse embedded information and context information and extract the semantic expression of the sentence through global maximum pooling. In the experimental part of the study, the research selected the Microsoft Research Paraphrase Corpus (MSRP) provided by Microsoft as the test data set and compared and analyzed the performance differences between the proposed method and existing algorithms. The results show that the proposed framework significantly improves the accuracy of translation. In summary, this study provides an effective solution to the problem of information distortion in real-time news translation.