Publications [#359826] of Lawrence Carin

Papers Published

1. 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 (January, 2020), pp. 9144-9156, ISBN 9781952148606
   (last updated on 2024/12/31)

   Abstract:
   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.