Multi-Aspect Controllable Text Generation with Disentangled Counterfactual Augmentation

Anonymous

Abstract: Multi-aspect controllable text generation aims to control the generated texts in attributes from multiple aspects (e.g., "positive" from sentiment and "sport'' from topic). Existing works neglect attribute correlations formed by the intertwining of different attributes. Particularly, the stereotype formed by imbalanced attribute correlations significantly affects multi-aspect control. In this paper, we propose MAGIC, a new multi-aspect controllable text generation method with disentangled counterfactual augmentation. We alleviate the issue of imbalanced attribute correlations during training using counterfactual feature vectors in the attribute latent space by disentanglement. During inference, we enhance attribute correlations by target-guided counterfactual augmentation to further improve multi-aspect control. Experiments show that MAGIC outperforms state-of-the-art baselines in both imbalanced and balanced attribute correlation scenarios.

Paper Type: long

Research Area: Generation

Contribution Types: NLP engineering experiment

Languages Studied: English

Preprint Status: There is no non-anonymous preprint and we do not intend to release one.

A1: yes

A1 Elaboration For Yes Or No: The "Limitations" section

A2: yes

A2 Elaboration For Yes Or No: The "Ethical Considerations" section

A3: yes

A3 Elaboration For Yes Or No: Abstract and Section 1

B: yes

B1: yes

B1 Elaboration For Yes Or No: Section 5.1; Appendices A and E

B2: yes

B2 Elaboration For Yes Or No: The "Ethical Considerations" section

B3: n/a

B4: n/a

B5: yes

B5 Elaboration For Yes Or No: Section 5.1

B6: yes

B6 Elaboration For Yes Or No: Section 5.1

C: yes

C1: yes

C1 Elaboration For Yes Or No: Appendix A

C2: yes

C2 Elaboration For Yes Or No: Section 5.1 and Appendix A

C3: yes

C3 Elaboration For Yes Or No: Section 5.2 and Appendices C and D

C4: n/a

D: yes

D1: n/a

D2: n/a

D3: n/a

D4: n/a

D5: n/a

E: yes

E1: n/a