Cross-domain sentiment classification via parameter transferring and attention sharing mechanism

Chuanjun Zhao, Suge Wang, Deyu Li, Xianzhi Liu, Xinyi Yang, Jinfeng Liu

Abstract: Highlights•Parameter transferring and attention sharing mechanism are proposed to effectively transfer sentiment knowledge through model transfer methods and effectively avoid overfitting.•Model transfer is presented toto transfer the network parameters from source domain network to target domain network.•Attention sharing mechanism is designed to share the attentional weights of different feature spaces in different domains.•The prediction results show the validity and advantages of sentiment transfer strategy.