In this paper, we implicitly incorporate morpheme information into word embedding. Based on the strategy we utilize the morpheme information, three models are proposed. To test the performances of our models, we conduct the word similarity and syntactic analogy. The results demonstrate