arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1712.08302cs.CL
收藏

用于神经标题生成的源端预测

Source-side Prediction for Neural Headline Generation

Tohoku University(东北大学);NTT Communication Science Laboratories, NTT Corporation(NTT通信科学实验室,NTT公司);Tokyo Institute of Technology(东京工业大学);RIKEN(理化学研究所)

Shun Kiyono, Sho Takase, Jun Suzuki, Naoaki Okazaki, Kentaro Inui, Masaaki Nagata

摘要:

编码器-解码器模型广泛用于自然语言生成任务中。然而,该模型有时会遭受重复的冗余生成,错过重要的短语,并包括不相关的实体。为了解决这些问题,我们提出了一种新的源端令牌预测模块。我们的方法联合估计源和目标词汇的概率分布,以捕获源和目标标记之间的对应关系。实验表明,该模型在标题生成任务中的性能优于当前最先进的方法。此外,我们表明,我们的方法有能力学习一个合理的令牌明智的对应关系,而不知道任何真正的路线。

英文摘要:

The encoder-decoder model is widely used in natural language generation tasks. However, the model sometimes suffers from repeated redundant generation, misses important phrases, and includes irrelevant entities. Toward solving these problems we propose a novel source-side token prediction module. Our method jointly estimates the probability distributions over source and target vocabularies to capture a correspondence between source and target tokens. The experiments show that the proposed model outperforms the current state-of-the-art method in the headline generation task. Additionally, we show that our method has an ability to learn a reasonable token-wise correspondence without knowing any true alignments.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑