Flexible entity marks and a fine-grained style control for knowledge based natural answer generation

作者：

Highlights：

• Random marks are used to compress the unnecessary semantic information of entities.

• Max-head mechanisms help extract more useful information.

• Fine-grained style control can generate answers of different styles.

摘要

•Random marks are used to compress the unnecessary semantic information of entities.•Max-head mechanisms help extract more useful information.•Fine-grained style control can generate answers of different styles.

论文关键词：Entity marks,Style control,Natural answer generation

论文评审过程：Received 22 October 2020, Revised 17 January 2022, Accepted 19 January 2022, Available online 10 February 2022, Version of Record 4 March 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108248