Integrating Semantic Scenario and Word Relations for Abstractive Sentence Summarization
31902Authors
nan
Abstract
AbstractRecently graph-based methods have been adopted for Abstractive Text Summarization. However, existing graph-based methods only consider either word relations or structure information, which neglect the correlation between them. To simultaneously capture the word relations and structure information from sentences, we propose a novel Dual Graph network for Abstractive Sentence Summarization. Specifically, we first construct semantic scenario graph and semantic word relation graph based on FrameNet, and subsequently learn their representations and design graph fusion method to enhance their correlation and obtain better semantic representation for summary generation. Experimental results show our model outperforms existing state-of-the-art methods on two popular benchmark datasets, i.e., Gigaword and DUC 2004.
Keywords
Full Content
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 2522ā2529 November 7ā11, 2021. cā2021 Association for Computational Linguistics 2522 2523 Filler Words Frames Sentence: Chinese vice premier huang ju said here Wednesday that worldwide engineers should cooperate with each other to contribute more to sustained development of the human society. Governed Leader premier Chinese huang ju Partner-1 Partner-2 cooperate with each other wordwide engineers Speaker Message said that society Giving Donor Theme Collaboration contribute Leadership Statement sustained development Intantionally_act F-to-F 2524 Chinese vice premier huang ju said ... Sentence BERT Encoder Word Relation Graph Semantic Scenario Graph Frame Embedding Feature Aggregator Graph Fusion CLS ... SEP Node Initialization Layer Node Initialization Layer Fusion Nodes Word Nodes Target Nodes FE Nodes Frame Nodes Multi-Head Attention Add & Norm Feed Forward Add & Norm Softmax Pvocab FE FE FE FE FE F F F F T T T W W W W W FE T W 2525 2526 2527 2528 2529
Labeled Text
AbstractRecently graph-based methods have been adopted for Abstractive Text Summarization. However, existing graph-based methods only consider either word relations or structure information, which neglect the correlation between them. To simultaneously capture the word relations and structure information from sentences, we propose a novel Dual Graph network for Abstractive Sentence Summarization. Specifically, we first construct semantic scenario graph and semantic word relation graph based on FrameNet, and subsequently learn their representations and design graph fusion method to enhance their correlation and obtain better semantic representation for summary generation. Experimental results show our model outperforms existing state-of-the-art methods on two popular benchmark datasets, i.e., Gigaword and DUC 2004.
Publication Details
- Paper ID:
- 31902
- Year:
- 2021
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan