Abstract: To address the problem of being unable to fully extract the relationship between text semantics and label in current multi-label text classification, a multi-label text classification model based on attention mechanism and convolutional neural network is proposed. The multi attention mechanism and CNN were used to represent the text, and the global and local semantic features of the text were fully mined. It combined tags and text information to calculate the interactive attention, and captured the relationship between tags after combining the text content. It used an adaptive fusion strategy to further extract the semantic information of the two. Experimental results show that this model can effectively improve the effect of multi label text classification compared with other mainstream models.