Skip to main content

Improving text generation with student-forcing optimal transport

Publication ,  Conference
Li, J; Li, C; Wang, G; Fu, H; Lin, YC; Chen, L; Zhang, Y; Tao, C; Zhang, R; Wang, W; Shen, D; Yang, Q; Carin, L
Published in: EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference
January 1, 2020

Neural language models are often trained with maximum likelihood estimation (MLE), where the next word is generated conditioned on the ground-truth word tokens. During testing, however, the model is instead conditioned on previously generated tokens, resulting in what is termed exposure bias. To reduce this gap between training and testing, we propose using optimal transport (OT) to match the sequences generated in these two modes. An extension is further proposed to improve the OT learning, based on the structural and contextual information of the text sequences. The effectiveness of the proposed method is validated on machine translation, text summarization, and text generation tasks.

Duke Scholars

Published In

EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference

Publication Date

January 1, 2020

Start / End Page

9144 / 9156
 

Citation

APA
Chicago
ICMJE
MLA
NLM
Li, J., Li, C., Wang, G., Fu, H., Lin, Y. C., Chen, L., … Carin, L. (2020). Improving text generation with student-forcing optimal transport. In EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference (pp. 9144–9156).
Li, J., C. Li, G. Wang, H. Fu, Y. C. Lin, L. Chen, Y. Zhang, et al. “Improving text generation with student-forcing optimal transport.” In EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference, 9144–56, 2020.
Li J, Li C, Wang G, Fu H, Lin YC, Chen L, et al. Improving text generation with student-forcing optimal transport. In: EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference. 2020. p. 9144–56.
Li, J., et al. “Improving text generation with student-forcing optimal transport.” EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference, 2020, pp. 9144–56.
Li J, Li C, Wang G, Fu H, Lin YC, Chen L, Zhang Y, Tao C, Zhang R, Wang W, Shen D, Yang Q, Carin L. Improving text generation with student-forcing optimal transport. EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference. 2020. p. 9144–9156.

Published In

EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference

Publication Date

January 1, 2020

Start / End Page

9144 / 9156