Complementary Auxiliary Classifiers for Label-Conditional Text Generation

By Yuan Li, Chunyuan Li, Yizhe Zhang, Xiujun Li, Guoqing Zheng, Lawrence Carin, Jianfeng Gao. 2020

23 citations recorded by the source corpus.

Learning to generate text with a given label is a challenging task because natural language sentences are highly variable and ambiguous. It renders difficulties in trade-off between sentence quality and label fidelity. In this paper, we present CARA to alleviate the issue, where two auxiliary classifiers work simultaneously to ensure that (1) the encoder learns disentangled features and (2) the generator produces label-related sentences. Two practical techniques are further proposed to improve the performance, including annealing the learning signal from the auxiliary classifier, and enhancing the encoder with pre-trained language models. To establish a comprehensive benchmark fostering future research, we consider a suite of four datasets, and systematically reproduce three representative methods. CARA shows consistent improvement over the previous methods on the task of label-conditional text generation, and achieves state-of-the-art on the task of attribute transfer.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Diyi Yang
2. Zhiting Hu
3. Xuezhi Wang
4. Yanzhe Zhang
5. Jiaao Chen
6. Olga Vechtomova
7. Di Jin
8. Rada Mihalcea

Citations in the corpus

• Continual Learning for Text Classification with Information Disentanglement Based Regularization (2021, background): “…text data, which has been extensively explored in generation tasks like style transfer (Fu et al., 2017; Zhao et al., 2018; Romanov et al., 2019; Li et al., 2020), where text hidden representations are often disentangled into sentiment (Fu et al., 2017; John et al., 2019), content (Romanov et…”
• Deep Learning for Text Style Transfer: A Survey (2020, methodology): “As the previous ACR explicitly requires the latent a to be classified by fc, AdvR trains from another perspective—enforcing that no attribute-related information is contained in z (Fu et al. 2018; Zhao et al. 2018; Romanov et al. 2019; John et al. 2019; Tikhonov et al. 2019; Li et al. 2020).”