The people's emotion reflect their attitude towards life and society. This paper proposes an approach to measure the positive emotion of Chinese comments from news portal and forums. A large emotion lexicon is first built by the Point mutual information (PMI) and expression symbols, and the method to calculate the emotion values of each comment is also provided based on this lexicon. Experiments on manual annotation dataset show that our method achieves important improvement compared with the

more »

... seline. Finally, we investigate the emotional differences among users' comments from different web sources, and find that the more strict the rules of the network platform, the more positive the user comments.