Air-Decoding: Attribute Distribution Reconstruction for Decoding-Time Controllable Text Generation

Tianqi Zhong, Quan Wang, Jingxuan Han, Yongdong Zhang, Zhendong Mao


Abstract
Controllable text generation (CTG) aims to generate text with desired attributes, and decoding-time-based methods have shown promising performance on this task. However, in this paper, we identify the phenomenon of Attribute Collapse for the first time. It causes the fluency of generated text to rapidly decrease when the control strength exceeds a critical value, rendering the text completely unusable. This limitation hinders the effectiveness of decoding methods in achieving high levels of controllability. To address this problem, we propose a novel lightweight decoding framework named Air-Decoding. Its main idea is reconstructing the attribute distributions to balance the weights between attribute words and non-attribute words to generate more fluent text. Specifically, we train prefixes by prefix-tuning to obtain attribute distributions. Then we design a novel attribute distribution reconstruction method to balance the obtained distributions and use the reconstructed distributions to guide language models for generation, effectively avoiding the issue of Attribute Collapse. Experiments on multiple CTG tasks prove that our method achieves a new state-of-the-art control performance.
Anthology ID:
2023.emnlp-main.512
Volume:
Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing
Month:
December
Year:
2023
Address:
Singapore
Editors:
Houda Bouamor, Juan Pino, Kalika Bali
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
8233–8248
Language:
URL:
https://aclanthology.org/2023.emnlp-main.512
DOI:
10.18653/v1/2023.emnlp-main.512
Bibkey:
Cite (ACL):
Tianqi Zhong, Quan Wang, Jingxuan Han, Yongdong Zhang, and Zhendong Mao. 2023. Air-Decoding: Attribute Distribution Reconstruction for Decoding-Time Controllable Text Generation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 8233–8248, Singapore. Association for Computational Linguistics.
Cite (Informal):
Air-Decoding: Attribute Distribution Reconstruction for Decoding-Time Controllable Text Generation (Zhong et al., EMNLP 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.emnlp-main.512.pdf
Video:
 https://aclanthology.org/2023.emnlp-main.512.mp4