Examining the impact of adaptive convolution on natural language understanding

作者：

Highlights：

• We propose an adaptive convolution for natural language understanding tasks.

• We design a set of filter-generating networks.

• Adaptive convolution provides greater flexibility to convolutional neural networks.

• We show the effectiveness of our approach through evaluation on nine datasets.

摘要

•We propose an adaptive convolution for natural language understanding tasks.•We design a set of filter-generating networks.•Adaptive convolution provides greater flexibility to convolutional neural networks.•We show the effectiveness of our approach through evaluation on nine datasets.

论文关键词：Natural language understanding,Adaptive convolution,Network adaptation

论文评审过程：Received 2 December 2019, Revised 24 July 2021, Accepted 4 October 2021, Available online 23 October 2021, Version of Record 3 November 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.116044