Year: 2019
Type: conference-paper
Abstract: In recent times, RNN-based sequence-tosequence attentional models have achieved good performance on abstractive summarization.However, numerous problems regarding repetition, incoherence, and exposure bias are encountered when applying these models.In this work, we propose a novel architecture that augments the standard sequence-to-sequence attentional model and a new training method combining reinforcement learning.We evaluate our proposed method on the CNN/Daily Mail dataset.The empirical results demonstrate the superiority of our proposed method in the abstractive summarization.
Authors: Chang, Yu, 99040, Lei, Hang, 99041, Li, Xiaoyu, 99042, Huang, Yiming, 99043
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: G6005326176, G7413685474
OpenAlex ID: W6996085020
