In this chapter, we will first summarize the content of this book, including classical and modern topic models, as well as how these models can be applied to solve practical tasks of emotion detection and aspect extraction. Then, we will point out the following promising directions for further investigation. (1) The enhancement of evaluation criteria: In evaluating flat topic models, the presence of polysemous words can significantly impact the coherence and diversity of topics, yet their influence is not fully captured by existing criteria. In evaluating hierarchical topic models, there is a notable lack of automated metrics specifically targeting the parent-child relationships within the topic hierarchy. (2) The integration of prior knowledge: Future work could explore integrating prior knowledge to guide the discovery process of topic structures, so as to improve both the interpretability and domain-specific applicability of the models. (3) The expansion of model capabilities: Considering the proliferation of multilingual and cross-domain data, it is valuable to construct a unified topic model capable of handling diverse text inputs. Finally, we expect more application scenarios to enhance the practical impact of topic models.