Cascading Broad Learning for Textual Emotion Classification

Although many DL-based methods are proposed for TEC task, they suffer from certain disadvantages such as long training time and difficult convergence. Motivated by the above challenges, and to improve the performance of general BL, in this chapter, we propose a new method for this task, named CBL, which is based on the cascading BL and MPNet. Texts are input into the MPNet to generate sentence embedding to better obtain semantic information. Cascading BL is utilized to improve the ability of feature extraction of text by cascading the feature nodes and enhancement nodes simultaneously in general BL. In addition, the L-curve method is applied to ensure balance between under-regularization and over-regularization in the regularization parameter optimization. Extensive experiments are conducted on two public data sets to verify the effectiveness.