Research map / Label-efficient and robust learning / Recognition and long-tail
Unsupervised Opinion Summarization with Content Planning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use aspect and sentiment content plans for synthetic sampling and summary generation | PlanSum · Yelp, Rotten Tomatoes, and Amazon | ROUGE-1/ROUGE-2/ROUGE-L: Yelp: 30.14 -> 34.79 (ROUGE-1), 5.26 -> 7.01 (ROUGE-2), 18.09 -> 19.74 (ROUGE-L); RT: 21.26 -> 21.77 (ROUGE-1), 4.61 -> | Helped |
| add aspect-sentiment disentanglement (ablation) | PlanSum content plan induction · Yelp, Rotten Tomatoes, and Amazon | ROUGE-L: Yelp: 18.83 -> 19.74; RT: 16.09 -> 16.98; Amazon: 18.52 -> 19.05 | Helped |
| add embedding uniqueness regularization (ablation) | PlanSum content plan induction · Yelp, Rotten Tomatoes, and Amazon | ROUGE-L: Yelp: 19.00 -> 19.74; RT: 16.85 -> 16.98; Amazon: 18.92 -> 19.05 | Helped |
| use content-plan-conditioned review sampling rather than random sampling (ablation) | PlanSum synthetic dataset creation · Yelp, Rotten Tomatoes, and Amazon | ROUGE-L: Yelp: 19.22 -> 19.74; RT: 16.61 -> 16.98; Amazon: 18.70 -> 19.05 | Helped |
| use content-plan-conditioned review sampling rather than lexical similarity sampling (ablation) | PlanSum synthetic dataset creation · Yelp, Rotten Tomatoes, and Amazon | ROUGE-L: Yelp: 19.38 -> 19.74; RT: 15.06 -> 16.98; Amazon: 18.31 -> 19.05 | Helped |
| use the content plan in the summarization model (ablation) | PlanSum · Yelp, Rotten Tomatoes, and Amazon | ROUGE-L: Yelp: 19.03 -> 19.74; RT: 16.56 -> 16.98; Amazon: 18.28 -> 19.05 | Helped |
| use injective fusion instead of mean token fusion (ablation) | PlanSum · Yelp, Rotten Tomatoes, and Amazon | ROUGE-L: Yelp: 18.72 -> 19.74; RT: 16.76 -> 16.98; Amazon: 18.57 -> 19.05 | Mixed |
| use BERT-based label smoothing prior instead of a uniform prior (ablation) | PlanSum · Yelp, Rotten Tomatoes, and Amazon | ROUGE-L: Yelp: 18.80 -> 19.74; RT: 16.77 -> 16.98; Amazon: 18.94 -> 19.05 | Helped |
Closest papers by meaning
- MeanSum: A Neural Model for Unsupervised Multi-Document Abstractive Summarization (ICML 2019 · similarity 0.89)
- Weakly-Supervised Opinion Summarization by Leveraging External Information (AAAI 2020 · similarity 0.89)
- PersonalSum: A User-Subjective Guided Personalized Summarization Dataset for Large Language Models (NeurIPS 2024 · similarity 0.85)
- Learning to Summarize by Learning to Quiz: Adversarial Agentic Collaboration for Long Document Summarization (ICLR 2026 · similarity 0.83)
- Unsupervised Extractive Summarization with Learnable Length Control Strategies (AAAI 2024 · similarity 0.82)
- Distilling Structured Rationale from Large Language Models to Small Language Models for Abstractive Summarization (AAAI 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.