ReadTwice: Reading Very Large Documents with Memories
Yury Zemlyanskiy, Joshua Ainslie, Michiel de Jong, Philip Pham, Ilya Eckstein, Fei Sha
Abstract
Knowledge-intensive tasks such as question answering often require assimilating information from different sections of large inputs such as books or article collections. We propose ReadTwice, a simple and effective technique that combines several strengths of prior approaches to model long-range dependencies with Transformers. The main idea is to read text in small segments, in parallel, summarizing each segment into a memory table to be used in a second read of the text. We show that the method outperforms models of comparable size on several question answering (QA) datasets and sets a new state of the art on the challenging NarrativeQA task, with questions about entire books.- Anthology ID:
- 2021.naacl-main.408
- Volume:
- Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
- Month:
- June
- Year:
- 2021
- Address:
- Online
- Venue:
- NAACL
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 5189–5195
- Language:
- URL:
- https://aclanthology.org/2021.naacl-main.408
- DOI:
- 10.18653/v1/2021.naacl-main.408
- Cite (ACL):
- Yury Zemlyanskiy, Joshua Ainslie, Michiel de Jong, Philip Pham, Ilya Eckstein, and Fei Sha. 2021. ReadTwice: Reading Very Large Documents with Memories. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 5189–5195, Online. Association for Computational Linguistics.
- Cite (Informal):
- ReadTwice: Reading Very Large Documents with Memories (Zemlyanskiy et al., NAACL 2021)
- PDF:
- https://preview.aclanthology.org/ingestion-script-update/2021.naacl-main.408.pdf
- Data
- HotpotQA, NarrativeQA, TriviaQA