A POMDP Dialogue Policy with 3-way Grounding and Adaptive Sensing for Learning through Communication
Abstract
Agents to assist with rescue, surgery, and similar activities could collaborate better with humans if they could learn new strategic behaviors through communication. We introduce a novel POMDP dialogue policy for learning from people. The policy has 3-way grounding of language in the shared physical context, the dialogue context, and persistent knowledge. It can learn distinct but related games, and can continue learning across dialogues for complex games. A novel sensing component supports adaptation to information-sharing differences across people. The single policy performs better than oracle policies customized to specific games and information behavior.- Anthology ID:
- 2022.findings-emnlp.504
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2022
- Month:
- December
- Year:
- 2022
- Address:
- Abu Dhabi, United Arab Emirates
- Editors:
- Yoav Goldberg, Zornitsa Kozareva, Yue Zhang
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 6767–6780
- Language:
- URL:
- https://aclanthology.org/2022.findings-emnlp.504
- DOI:
- 10.18653/v1/2022.findings-emnlp.504
- Cite (ACL):
- Maryam Zare, Alan Wagner, and Rebecca Passonneau. 2022. A POMDP Dialogue Policy with 3-way Grounding and Adaptive Sensing for Learning through Communication. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 6767–6780, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
- Cite (Informal):
- A POMDP Dialogue Policy with 3-way Grounding and Adaptive Sensing for Learning through Communication (Zare et al., Findings 2022)
- PDF:
- https://preview.aclanthology.org/ingest-acl-2023-videos/2022.findings-emnlp.504.pdf