Tibetan is a language with very flexible word order. Current research such as Tibetan lexical analysis and syntactic analysis cannot meet the needs of Tibetan natural language understanding. Based on the simple sentence pattern of Tibetan sentences, we study the semantic dependence of Tibetan based on projection, construct a Tibetan semantic dependency tree library, and design a semantic dependency arc type analysis feature template. Finally, the semantically dependent arc sentences of the artificial analysis are analyzed and labeled through the maximum entropy classification model, which provides a new perspective and sound theoretical support for future semantic dependence analysis.