Continual Learning for Grounded Instruction Generation by Observing Human Following Behavior
AbstractWe study continual learning for natural language instruction generation, by observing human users’ instruction execution. We focus on a collaborative scenario, where the system both acts and delegates tasks to human users using natural language. We compare user execution of g...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
The MIT Press
2021-01-01
|
| Series: | Transactions of the Association for Computational Linguistics |
| Online Access: | https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00428/108610/Continual-Learning-for-Grounded-Instruction |
| _version_ | 1818215423887278080 |
|---|---|
| author | Noriyuki Kojima Alane Suhr Yoav Artzi |
| author_facet | Noriyuki Kojima Alane Suhr Yoav Artzi |
| author_sort | Noriyuki Kojima |
| collection | DOAJ |
| description |
AbstractWe study continual learning for natural language instruction generation, by observing human users’ instruction execution. We focus on a collaborative scenario, where the system both acts and delegates tasks to human users using natural language. We compare user execution of generated instructions to the original system intent as an indication to the system’s success communicating its intent. We show how to use this signal to improve the system’s ability to generate instructions via contextual bandit learning. In interaction with real users, our system demonstrates dramatic improvements in its ability to generate language over time. |
| first_indexed | 2024-12-12T06:35:51Z |
| format | Article |
| id | doaj.art-df09a249402a44debc6aed2c01716cea |
| institution | Directory Open Access Journal |
| issn | 2307-387X |
| language | English |
| last_indexed | 2024-12-12T06:35:51Z |
| publishDate | 2021-01-01 |
| publisher | The MIT Press |
| record_format | Article |
| series | Transactions of the Association for Computational Linguistics |
| spelling | doaj.art-df09a249402a44debc6aed2c01716cea2022-12-22T00:34:29ZengThe MIT PressTransactions of the Association for Computational Linguistics2307-387X2021-01-0191303131910.1162/tacl_a_00428Continual Learning for Grounded Instruction Generation by Observing Human Following BehaviorNoriyuki Kojima0Alane Suhr1Yoav Artzi2Department of Computer Science and Cornell Tech, Cornell University, USA. nk654@cornell.eduDepartment of Computer Science and Cornell Tech, Cornell University, USA. suhr@cs.cornell.eduDepartment of Computer Science and Cornell Tech, Cornell University, USA. yoav@cs.cornell.edu AbstractWe study continual learning for natural language instruction generation, by observing human users’ instruction execution. We focus on a collaborative scenario, where the system both acts and delegates tasks to human users using natural language. We compare user execution of generated instructions to the original system intent as an indication to the system’s success communicating its intent. We show how to use this signal to improve the system’s ability to generate instructions via contextual bandit learning. In interaction with real users, our system demonstrates dramatic improvements in its ability to generate language over time.https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00428/108610/Continual-Learning-for-Grounded-Instruction |
| spellingShingle | Noriyuki Kojima Alane Suhr Yoav Artzi Continual Learning for Grounded Instruction Generation by Observing Human Following Behavior Transactions of the Association for Computational Linguistics |
| title | Continual Learning for Grounded Instruction Generation by Observing Human Following Behavior |
| title_full | Continual Learning for Grounded Instruction Generation by Observing Human Following Behavior |
| title_fullStr | Continual Learning for Grounded Instruction Generation by Observing Human Following Behavior |
| title_full_unstemmed | Continual Learning for Grounded Instruction Generation by Observing Human Following Behavior |
| title_short | Continual Learning for Grounded Instruction Generation by Observing Human Following Behavior |
| title_sort | continual learning for grounded instruction generation by observing human following behavior |
| url | https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00428/108610/Continual-Learning-for-Grounded-Instruction |
| work_keys_str_mv | AT noriyukikojima continuallearningforgroundedinstructiongenerationbyobservinghumanfollowingbehavior AT alanesuhr continuallearningforgroundedinstructiongenerationbyobservinghumanfollowingbehavior AT yoavartzi continuallearningforgroundedinstructiongenerationbyobservinghumanfollowingbehavior |