In this paper,a comparative analysis is conducted between machine translation and human translation in the 2 500 linguistic data which are randomly selected from a well-known auto brand's manual in Automobile Document Translation Corpus of Automotive College,Tongji University.In addition,420 sentences including the English connective "and" are further analyzed.Common mistranslation types are summarized and logical semantic relations are discussed in the paper,hoping to provide some insights on upgrade of post-editing system for machine translation.It is found that accurate translation of connective "and" not only need to consider syntax,but also semantics.Only in this way can machine generate more accurate translations.