With the continuous growth of modern archive management data, it is pretty important to effectively partition the archive texts into classes. Therefore, this paper proposes two incremental multi-modal text data clustering methods. By multi-perspective analysis of the text content, the potential topic features of texts are integrated to improve the accuracy of text clustering. In addition, the corresponding incremental multi-modal feature learning models for text clustering are designed to improve the efficiency of massive and dynamic text partition. The experimental results on two real-world text data sets show that the proposed incremental multimodal text clustering methods in this paper have superior performance and can effectively classify text data.