Abstract: 【Purpose/significance】Emotion recognition in conversations is a burgeoning and vital subfield within sentiment analysis. It is devoted to accurately discerning each utterance′s emotional state from an array of predefined emotions. It applies to not only senti⁃ ment mining of emotion-focused social media but also real-time dialogue emotion analysis, positively impacting sectors such as public sentiment analysis and user emotion insight【. Method/process】This paper presents a dialogue emotion recognition model named ERCPrKE. The model integrates knowledge gleaned from external structured data, offering additional contextual information for dialogue comprehension. Such integration boosts the modeling of dialogues and enriches the representation of context. Moreover, we incorpo⁃ rated a prompt template after the target dialogue input to help the model distinguish between the target utterance and its context more effectively.【Result/conclusion】The experimental outcomes suggest that the performance of dialogue emotion recognition can be en⁃ hanced by implementing prompts and introducing external knowledge. The performance has improved by over 2% on the MELD, Emo⁃ ryNLP, and IEMOCAP datasets, with a significant leap of 4.05% on the EmoryNLP dataset【. Innovation/limitation】We have proposed a dialogue sentiment analysis model that integrates prompts and external knowledge, paving a fresh path to enhancing users′ emotional experience amid the swift advancement of dialogic artificial intelligence. Future research will consider integrating various sources of external knowledge to bolster the robustness and precision of emotion identification.