Abstract
Transformer-based pre-trained language models boost the performance of open-domain dialogue systems. Prior works leverage Transformer-based pre-trained language models to generate texts with desired attributes in two general approaches: (1) gradient-based methods: updating all latent representations of pre-trained models with gradients from attribute models; (2) weighted-decoding methods: re-ranking beam candidates from pre-trained models with attribute functions. However, gradient-based methods lead to high computation cost and can easily get overfitted on small training sets, while weighted-decoding methods are inherently constrained by the low-variance high-bias pre-trained model. In this work, we propose a novel approach to control the generation of Transformer-based pre-trained language models: the SideControl framework, which leverages a novel control attributes loss to incorporate useful control signals, and is shown to perform well with very limited training samples. We evaluate our proposed method on two benchmark open-domain dialogue datasets, and results show that the SideControl framework has better controllability, higher generation quality and better sample-efficiency than existing gradient-based and weighted-decoding baselines.- Anthology ID:
- 2021.findings-emnlp.188
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2021
- Month:
- November
- Year:
- 2021
- Address:
- Punta Cana, Dominican Republic
- Editors:
- Marie-Francine Moens, Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
- Venue:
- Findings
- SIG:
- SIGDAT
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 2175–2194
- Language:
- URL:
- https://aclanthology.org/2021.findings-emnlp.188
- DOI:
- 10.18653/v1/2021.findings-emnlp.188
- Cite (ACL):
- Wanyu Du and Yangfeng Ji. 2021. SideControl: Controlled Open-domain Dialogue Generation via Additive Side Networks. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 2175–2194, Punta Cana, Dominican Republic. Association for Computational Linguistics.
- Cite (Informal):
- SideControl: Controlled Open-domain Dialogue Generation via Additive Side Networks (Du & Ji, Findings 2021)
- PDF:
- https://preview.aclanthology.org/proper-vol2-ingestion/2021.findings-emnlp.188.pdf
- Code
- wyu-du/controlled-dialogue-generation
- Data
- ConvAI2, DailyDialog