We　propose　a　language　model　of　mix　CNN　(Convolution　Neural　Network)　with　bi-RNN　(Bi-directional　Recurrent　Neural　Network)　to　classify　the　text　at　the　character-level.　Unlike　word-level　model　is　that　avoiding　the　problem　of　unregistered　words　and　improves　the　robustness　of　the　text　representation　in　character-level　model.　The　language　model　mainly　uses　the　data　augment　by　different　convolution　filters　of　CNN　and　then　the　bi-RNN　obtain　the　contextual　information　in　both　directions　to　classify　the　text.　The　results　show　that　this　model　have　a　better　performance　than　the　common　CNN　and　LSTM(long　short-term　memory)　classification　methods.　©　2018　IEEE.