Abstract: This paper introduces the development of rule-based machine translation, statistical machine translation and neural machine translation, and illustrates that machine translation research needs to be supported by linguistic knowledge and common sense, and proposes that the currently widely popular connectionist approach based on linguistic big data should not be overly fetishized, and the symbolist approach based on linguistic rules and common sense, which is currently receiving a lukewarm reception, should not be easily ignored.The connectionist approach based on linguistic big data and the symbolist approach based on linguistic rules and common sense should be skillfully and precisely combined, so as to push the research of machine translation into depth.This paper points out that machine translation will become a good friend and powerful assistant of human translation, and that machine translation and human translation should boost each other.