In this chapter, we bridge emotion theories with the recent advancements in Natural Language Processing (NLP). Central to our discussion is the theoretical debate between categorical and dimensional emotion theories, exploring whether emotions are recognized as innate, discrete entities or as dynamic, context-dependent constructs. We also highlight the critical but often oversimplified role of emotion valence (positive or negative emotions) in emotion detection. Regarding NLP, we explore how emotion detection algorithms use categorical and dimensional models of emotion to interpret texts, emphasising the need for nuanced, multidimensional approaches to detect emotions in various applications, including chatbots, social media, review systems, literary studies, mental health, and humour, sarcasm and hate speech detection. This overview underscores emotion detection’s interdisciplinary and evolving nature, balancing theoretical insights with practical applications in online communication.