Unlike English and other western languages, there are no delimiters to mark word boundaries in both Chinese and Tibetan. Therefore, word segmentation is the first step for Chinese and Tibetan natural language processing such as machine translation and information retrieval. However, Chinese word segmentation has been studied for many years and the technology is relatively mature. In contrast, Tibetan word segmentation is less concerned by researchers. In this paper, we learn from Chinese word segmentation approach and analysis the characteristic of Tibetan language, designs a Tibetan word segmentation system based on conditional random fields. The experiment shows that the algorithm is effective and can be preliminary applied.