Reproduce “Improving Document-Level Sentiment Classification Using Importance of Sentences” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Gihyeon Choi, Shinhyeok Oh, Harksoo Kim
Researchers at Kangwon National University and Konkuk University developed a deep neural network model that explicitly quantifies and utilizes the 'importance degree' of each sentence for document-level sentiment classification. This approach consistently achieved higher accuracy compared to strong baselines across multiple datasets by focusing on the most sentiment-salient sentences.
Key insights
- Empirically validated that considering the varying importance of sentences within a document significantly enhances the accuracy of document-level sentiment classification.
- The proposed deep neural network architecture successfully integrates a novel gate mechanism to determine and incorporate sentence importance, leading to superior performance.
- The inclusion of 'class similarity embeddings' at both sentence and document levels contributes to more informed sentiment representations and improved classification.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.