Implicit emotional tendency recognition is a challenging task, for the patterns of implicit emotional sentences have no obvious sentiment words or privative words. Currently, deep learning approaches have shown a huge potential for this problem. This paper proposes a hierarchical disconnected recurrent neural network to overcome the problem of the lack of emotional information in implicit sentiment sentence recognition. The proposed network first encodes both words and characters in implicit sentiment sentence by using the disconnected recurrent neural network and then fuses the context information of this sentence through the hierarchical structure. The capsule network is adopted to construct different fine-grained context information for extracting high-level feature information and provide additional semantic information for emotion recognition by using the context information. This approach improves the accuracy in implicit emotion recognition. Experimental results prove that the proposed model outperform various current mainstream models. The F 1 of ours reaches 81.5%, which is 2%-3% higher than those of the current mainstream models.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.