Conversations depend on information from the context. To go beyond one-round conversation, a chatbot must resolve contextual information such as:1) co-reference resolution, 2) ellipsis resolution,and 3) conjunctive relationship resolution.There are simply not enough data to avoid these problems by trying to train a sequence-to-sequencemodel for multi-round conversation similar to that of one-round conversation.The contributions of this paper are: 1) We formulate the problem ofcontext resolution for conversation; 2) We present deep learning models, includingan end-to-end network for context resolution; 3) We propose a way of creating a huge amount of