Personalized Implicit Sentiment Analysis Based on Multi-view Fusion of Implicit User Preference

Implicit sentiment analysis is one of the challenges in the field of sentiment analysis. Implicit sentiment lacks explicit sentimental words and have subjective differences. Previous methods have mostly not deeply analyzed user information, resulting in an inability to achieve accurate personalized implicit sentiment analysis. In this study, we propose a personalized implicit sentiment analysis model based on multi-view fusion of implicit user preference. We use the interaction relationship between users and words to construct the user-content interaction graph. By generating multiple views and learning them separately, we obtain implicit user preferences from different perspectives. Meanwhile, we also model the content knowledge of users. Finally, we integrate user interaction knowledge and content knowledge into implicit sentiment sentence representation. We conducted experiments on two datasets, and the results showed that our model outperforms other baseline models in terms of performance.