With the rapid application of social media, hot topic detection is an essential issue on mining content and public opinion analysis. As some traditional topic models, such as probability latent semantic analysis (PLSA), latent dirichlet allocation (LDA), cannot mine the short texts effectively for the serious sparse problems, this paper proposes a micro-blog topic detection model based on the semantic dependency parsing. The proposed model parses micro-blog by semantic dependency parsing and analyzes the effect of semantic dependency relationships on topic detection. Furthermore, according to the previous works on analyzing and removing the irrelevant semantic dependency relationships for topic detection, this model makes use of the advantages of the BTM, and then combines the traditional BTM with semantic dependency relationships. The evaluation shows that the proposed model can obtain a more coherent topic in topic detection.