Entity discovery and linking can help to understand the questions semantics and infer answers in question answering systems. In this paper, we study the characteristics of Chinese questions and propose a joint
framework that leverages the mutual dependency between entity discovery and linking to enhance their performances. It jointly connects a joint parsing method based on concept knowledge tree for entity discovery, with candidate entity generation of entity linking. And we also investigate conditional random fields to detect entity mentions and filter
them with candidate entity generation. Experiments show that our proposed method outperforms state-of-the-art methods on a real dataset.