This paper presents an emotion recognition system with textual input. In this system, an emotional semantic network is proposed to extract the semantic information related to emotion. The semantic network is composed of two subnetworks: a static semantic network and a dynamic semantic network. The static semantic network is established from an existing Chinese knowledge base called HowNet and used to estimate the emotion trigger value of each word. The dynamic semantic network accepts the textual input and dynamically constructs the nodes and links, which represent the emotion carrier and the emotion propagator respectively. Initiated by the emotion trigger value, the emotion in the dynamic semantic network will propagate and finally converge to the final emotion output. Experimental results show an encouraging achievement was obtained.