Hierarchical Interaction Networks with Rethinking Mechanism for Document-Level Sentiment Analysis

Document-level Sentiment Analysis (DSA) is more challenging due to vague semantic links and complicate sentiment information. Recent works have been devoted leveraging text summarization achieved promising results. However, these summarization-based methods did not take full advantage of the summary including ignoring inherent interactions between document. As a result, they limited representation express major points in document, which highly indicative key sentiment. In this paper, we study how effectively generate discriminative with explicit subject patterns contexts for DSA. A Hierarchical Interaction Networks (HIN) proposed explore bidirectional document at multiple granularities learn subject-oriented representations classification. Furthermore, design Sentiment-based Rethinking mechanism (SR) by refining HIN label information sentiment-aware representation. We extensively evaluate our models on three public datasets. The experimental results consistently demonstrate effectiveness show that HIN-SR outperforms various state-of-the-art methods.