Dual Broad Learning for Textual Emotion Classification

Although BBL and CBL were proposed in Chaps. 2 and 3 , respectively, and they have achieved good results, there is room for improvement of performance of TEC task. Thus, in this chapter, we propose a new method for TCE task, named DBL, which is based on dual BL. DBL consists of two components like BL-BERT and BL-MPNet. As to BL-BERT, texts are input into BERT to generate sentence embedding containing useful information and then the output is input into BL. As to BL-MPNet, texts are input into MPNet to generate sentence embedding containing useful information, and then the output is also input into BL. Finally, it needs to conduct the concatenation for fusing the features from BERT and MPNet.