Research on Improving the Performance of English Machine Translation Through Optimal Extraction of Language Vector Features

This article briefly introduces the Word2vec model used for text vectorization and the convolutional neural network (CNN) extractor used for extracting semantic features. The CNN extractor was applied in machine translation, and the semantic feature vector was combined with the intermediate vector of the encoder to optimize the translation algorithm. Simulation experiments were conducted to compare this algorithm with the recurrent neural network and traditional long short-term memory (LSTM) models. The results showed that the machine translation algorithm adopting Word2vec performed better than that adopting one-hot encoding and was less affected by sentence length. The CNN extractor effectively extracted semantic features. The LSTM algorithm that combined semantic features not only was less affected by sentence length but also achieved higher accuracy in translations, resulting in more fluent translations.