| 英文摘要 |
The objective of Sentence-level Revision (SentRev) is to enhance the fluency of English writing; however, the performance of the three baseline methods is notably suboptimal. In this study, we propose a method utilizing neural reinforcement learning, tailored to the specific characteristics of this task, which has resulted in superior performance over the baseline methods, surpassing them in multiple evaluation metrics. Moreover, we have identified conspicuous bottlenecks in SentRev’s efficacy in improving the fluency of English writing. |