File Download
There are no files associated with this item.
Supplementary
-
Citations:
- Appears in Collections:
Conference Paper: An Exploratory Study on Long Dialogue Summarization: What Works and What's Next
Title | An Exploratory Study on Long Dialogue Summarization: What Works and What's Next |
---|---|
Authors | |
Issue Date | 2021 |
Publisher | Association for Computational Linguistics. |
Citation | The 2021 Conference on Empirical Methods in Natural Language Processing, Findings of the ACL: EMNLP 2021, p. 4426–4433 How to Cite? |
Abstract | Dialogue summarization helps readers capture salient information from long conversations in meetings, interviews, and TV series. However, real-world dialogues pose a great challenge to current summarization models, as the dialogue length typically exceeds the input limits imposed by recent transformer-based pre-trained models, and the interactive nature of dialogues makes relevant information more context-dependent and sparsely distributed than news articles. In this work, we perform a comprehensive study on long dialogue summarization by investigating three strategies to deal with the lengthy input problem and locate relevant information: (1) extended transformer models such as Longformer, (2) retrieve-then-summarize pipeline models with several dialogue utterance retrieval methods, and (3) hierarchical dialogue encoding models such as HMNet. Our experimental results on three long dialogue datasets (QMSum, MediaSum, SummScreen) show that the retrieve-then-summarize pipeline models yield the best performance. We also demonstrate that the summary quality can be further improved with a stronger retrieval model and pretraining on proper external summarization datasets. |
Persistent Identifier | http://hdl.handle.net/10722/319364 |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Zhang, Y | - |
dc.contributor.author | Ni, A | - |
dc.contributor.author | Yu, T | - |
dc.contributor.author | Zhang, R | - |
dc.contributor.author | Zhu, C | - |
dc.contributor.author | Deb, B | - |
dc.contributor.author | Celikyilmaz, A | - |
dc.contributor.author | Awadallah, A | - |
dc.contributor.author | Radev, D | - |
dc.date.accessioned | 2022-10-14T05:11:58Z | - |
dc.date.available | 2022-10-14T05:11:58Z | - |
dc.date.issued | 2021 | - |
dc.identifier.citation | The 2021 Conference on Empirical Methods in Natural Language Processing, Findings of the ACL: EMNLP 2021, p. 4426–4433 | - |
dc.identifier.uri | http://hdl.handle.net/10722/319364 | - |
dc.description.abstract | Dialogue summarization helps readers capture salient information from long conversations in meetings, interviews, and TV series. However, real-world dialogues pose a great challenge to current summarization models, as the dialogue length typically exceeds the input limits imposed by recent transformer-based pre-trained models, and the interactive nature of dialogues makes relevant information more context-dependent and sparsely distributed than news articles. In this work, we perform a comprehensive study on long dialogue summarization by investigating three strategies to deal with the lengthy input problem and locate relevant information: (1) extended transformer models such as Longformer, (2) retrieve-then-summarize pipeline models with several dialogue utterance retrieval methods, and (3) hierarchical dialogue encoding models such as HMNet. Our experimental results on three long dialogue datasets (QMSum, MediaSum, SummScreen) show that the retrieve-then-summarize pipeline models yield the best performance. We also demonstrate that the summary quality can be further improved with a stronger retrieval model and pretraining on proper external summarization datasets. | - |
dc.language | eng | - |
dc.publisher | Association for Computational Linguistics. | - |
dc.relation.ispartof | The 2021 Conference on Empirical Methods in Natural Language Processing | - |
dc.title | An Exploratory Study on Long Dialogue Summarization: What Works and What's Next | - |
dc.type | Conference_Paper | - |
dc.identifier.email | Yu, T: taoyds@hku.hk | - |
dc.identifier.authority | Yu, T=rp02864 | - |
dc.identifier.hkuros | 339279 | - |
dc.identifier.volume | Findings of the ACL: EMNLP 2021 | - |
dc.identifier.spage | 4426–4433 | - |
dc.identifier.epage | 4426–4433 | - |