As the rapid development of social media, now massive data are produced every day. We need to know about users' view and comment on a hot event so that the government and relevant departments can control and guide public opinion, and people can have a correct understanding of the event. So it is necessary to analyze these data. While it is difficult to dig out sufficient information from these data because of their shortness and sparsity. In this paper, we propose a sentiment orientation analysis method based on background and domain sentiment lexicon expansion, which reflects user's sentiment orientation of evaluation objects more comprehensively. Then we can obtain a more informative and significant sentiment orientation of short text. In the experiment, Chinese Weibo data sets of French security and Turkish Airlines are used to show that our sentiment orientation analysis method has better performance than the domain sentiment lexicon based method.