📄Paper : Fast Abstractive Summarization with Reinforce-Selected Sentence Rewriting
💡 강화 학습을 사용하여 Extractive Abstractive model을 연결한 end-to-end 프레임워크
word-sentence hierarchical framework
→ Sentence level의 extract를 수행한 후, word-level의 rewrite 수행
💡 Contribution 정리
1. sentence-level policy gradient method (RL)
sentence-level의 Extractor와 word-level의 Abstractor를 연결함으로써
word-sentence hierarchy 프레임워크 구현
→ 언어 구조를 모델링하는 데 효과적이며, 병렬화 (parallel decoding) 를 가능하게 함
2. 모델 속도 개선
extract와 rewrite이 병렬적으로 동작하는 parallel decoding로 인해 모델 속도 개선
inference speed 10-20배 개선, training speed 4배 개선
🌱 세미나 자료
👇 세미나 자료 보러가기
✨ 리뷰한 논문
Chen, Y. C. and Bansal, M. "Fast abstractive summarization with reinforce-selected sentence rewriting." Proc. of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Melbourne, Australia, pp. 675-686, 2018.
✨ 참고 문헌
-
Huang, S., Wang, R., Xie, Q., Li, L., and Liu, Y. "An extraction-abstraction hybrid approach for long document summarization." Proc. of the 2019 6th International Conference on Behavioral, Economic and Socio-Cultural Computing (BESC), Beijing, China, pp. 1-6, 2019.
-
Vinyals, O., Fortunato, M., and Jaitly, N. "Pointer networks.“ Proc. of the 28th International Conference on Neural Information Processing Systems - Volume 2, Montreal, Canada, pp. 2692-2700, 2015.
-
Vinyals, O., Bengio, S., and Kudlur, M. "Order matters: sequence to sequence for sets.“ Proc. of the 4th International Conference on Learning Representations, San Juan, Puerto Rico, 2016.
