Research map / Graph learning / Knowledge and temporal graphs
SemSUM: Semantic Dependency Guided Neural Abstractive Summarization
Hanqi Jin, Tianming Wang, Xiaojun Wan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add semantic dependency graph encoder | Transformer · Gigaword test | ROUGE-2 F1: 18.08 -> 19.75 | Improved |
| add semantic dependency graph encoder | Transformer · Gigaword test | BERTScore F1: 60.48 -> 61.56 | Improved |
| add semantic dependency graph encoder | Transformer · DUC2004 test | ROUGE-1 recall: 29.78 -> 31.00 | Improved |
| add semantic dependency graph encoder | Transformer · MSR-ATC test | ROUGE-2 F1: 12.45 -> 17.08 | Improved |
| graph attention encoder instead of GCN | TFM&GCN · Gigaword test | ROUGE-2 F1: 19.03 -> 19.75 | Improved |
| add graph encoder to sentence encoder (ablation) | SemSUM · Gigaword dev | ROUGE-2 F1: 24.94 -> 26.54 | Improved |
| add sentence encoder to graph encoder (ablation) | SemSUM · Gigaword dev | ROUGE-2 F1: 25.21 -> 26.54 | Improved |
| add layer aggregation (BiLSTM) (ablation) | SemSUM · Gigaword dev | ROUGE-2 F1: 25.83 -> 26.54 | Improved |
| separate in/out direction mappings (ablation) | SemSUM · Gigaword dev | ROUGE-2 F1: 25.57 -> 26.54 | Improved |
| add position-adjacent relations as edge type to graph-only encoder (ablation) | SemSUM graph-only · Gigaword dev | ROUGE-2 F1: 25.21 -> 25.51 | Improved |
| SemSUM vs BiSET | SemSUM · Gigaword test | ROUGE-2 F1: 19.78 vs 19.75 | Mixed |
| add semantic graph | Transformer · Gigaword human eval | informativeness: 3.27 -> 3.8 | Improved |
Later papers that compared with this method
1 later accepted paper reports SemSUM as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Hierarchical Heterogeneous Graph Attention Network for Syntax-Aware Summarization (AAAI 2022 · better on 4 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Joint Parsing and Generation for Abstractive Summarization (AAAI 2020 · similarity 0.88)
- MeanSum: A Neural Model for Unsupervised Multi-Document Abstractive Summarization (ICML 2019 · similarity 0.88)
- Summarization Programs: Interpretable Abstractive Summarization with Neural Modular Trees (ICLR 2023 · similarity 0.88)
- Controlling the Amount of Verbatim Copying in Abstractive Summarization (AAAI 2020 · similarity 0.88)
- Structured Neural Summarization (ICLR 2019 · similarity 0.87)
- Distilling Structured Rationale from Large Language Models to Small Language Models for Abstractive Summarization (AAAI 2025 · similarity 0.86)
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.
