Topic discovery and emotion detection are closely related such that public emotions always vary from one topic to another, and topics trigger public emotions. For instance, topics with representative words such as “care,” “death,” and “discover” are unvaried clues that, respectively, evoke emotions of “touching,” “sadness,” and “surprise” in users. However, there are also topics that may trigger different emotions among user groups as characterized by stance, gender, region, education, and others. This phenomenon is quite common in examples such as after a football match ends, the match result will evoke different emotions among fans from the two opposing sides. In light of this consideration, we propose a method of topic discovery and emotion detection with the help of user characteristics. To reduce noise in abundant features, we further propose two fast supervised topic models that are equipped with an accelerated algorithm. Finally, we develop an aspect extraction approach guided by document-topic distributions derived from neural topic models. This chapter will describe our work on topic-level emotion detection and topic-aware aspect extraction. For the former, we introduce topic-level emotion models exploiting user groups and labels, respectively. For the latter, we present a topic-aware dynamic convolutional neural network for aspect extraction.