Selective Self-Attention Networks

This repo provides the code for ACL 2020 paper How Does Selective Mechanism Improve Self-attention Networks ?. The code is based on THUMT and Tensorflow 1.6. Please cite our paper if you find this repository helpful in your research:

@inproceedings{geng2020selective,
    title = "How Does Selective Mechanism Improve Self-Attention Networks?",
    author = "Geng, Xinwei  and Wang, Longyue  and Wang, Xing  and Qin, Bing  and Liu, Ting  and Tu, Zhaopeng",
    booktitle = "ACL",
    month = jul,
    year = "2020",
}