Due to the diversity and complexity of irony as a linguistic phenomenon, irony detection has always been a challenging research field in text sentiment analysis, which helps to identify ironic tones and accurately judge the sentiment of the text. In this paper, we discuss the research status quo of irony detection in the field of NLP and outlines various existing irony detection models firstly, and then we propose the LADCCL model for Chinese irony detection in social media. In LADCCL model, we try to process labels as textual information as well, and increase the semantic interaction between words and labels through the interaction computation of label embedding. Meanwhile, for the problem of fuzzy definition of ironic text, a label smoothing mechanism is added to allow the model to improve the perception of all classifications through parameter perturbation. Meanwhile, we find that the traditional binary classification of irony and non-irony is not sufficient to distinguish various types of irony well. Thus, we design another Chinese sub-dataset for four-class classification.