Machine Translation System Based on Semantic Selection and Information Features

In recent years, corpus-based machine translation (MT) has received widespread attention from scholars, among which machine translation based on semantic selection and information features has also entered people’s vision. This article intended to start from the perspective of machine translation and use semantic selection models to increase attention to semantics and solve the problems of mistranslation and omission caused by semantic neglect. On this basis, this article intended to study cross-language and cross-modal semantic space learning methods based on information features, improve the semantic extraction ability of the encoding end, and decompose them into machine translation models. Each sentence was set to consist of 10 Chinese characters, and Chinese was translated into English. When the number of sentences was 4, the accuracy of MT systems based on semantic selection and information features was 99.4%, while the accuracy of MT systems based on semantic selection was 98.3%. A MT framework based on semantic selection and information features can organically combine sentence structure and semantic information and better focus on semantic vocabulary in sentences, thereby reducing mistranslation and improving translation quality.