The paper introduces DailyDialog, a high-quality, manually labeled multi-turn dialogue dataset designed to advance research in dialog systems. The dataset reflects real-life communication patterns and covers a wide range of daily topics, including relationships, ordinary life, and work. It is characterized by its formal language, specific dialog flow patterns, and rich emotional content. The dataset is constructed by crawling English learning resources and includes 13,118 dialogues with an average of 8 speaker turns per dialogue. The authors manually label the dataset with communication intentions and emotions, aiming to provide a comprehensive resource for evaluating existing approaches in dialog systems. The paper also presents an evaluation of retrieval-based and generation-based approaches on the dataset, highlighting the effectiveness of incorporating intention and emotion information in response retrieval and generation tasks.The paper introduces DailyDialog, a high-quality, manually labeled multi-turn dialogue dataset designed to advance research in dialog systems. The dataset reflects real-life communication patterns and covers a wide range of daily topics, including relationships, ordinary life, and work. It is characterized by its formal language, specific dialog flow patterns, and rich emotional content. The dataset is constructed by crawling English learning resources and includes 13,118 dialogues with an average of 8 speaker turns per dialogue. The authors manually label the dataset with communication intentions and emotions, aiming to provide a comprehensive resource for evaluating existing approaches in dialog systems. The paper also presents an evaluation of retrieval-based and generation-based approaches on the dataset, highlighting the effectiveness of incorporating intention and emotion information in response retrieval and generation tasks.