This is the official repository of the code and data of the paper Automatic Construction of Sememe Knowledge Bases via Dictionaries, which is published on Findings of ACL: ACL-IJCNLP 2021 [pdf].
-
Download Glove vectors, which is used for NLI, CR and Textual Adversarial Attack experiments.
-
Download Stanford POS Tagger, which is used for generating candidate words for Textual Adversarial Attack. However, you can directly use our preprocessed data.
Please see the README.md
files in ConsistencyCheck/
, CR/
, LM_SDLM/
and NLI/
for specific running instructions for each model on corresponding intrinsic evaluation and downstream tasks.
Please kindly cite our paper:
@article{qi2021automatic,
title={Automatic Construction of Sememe Knowledge Bases via Dictionaries},
author={Qi, Fanchao and Chen, Yangyi and Wang, Fengyu and Liu, Zhiyuan and Chen, Xiao and Sun, Maosong},
journal={arXiv preprint arXiv:2105.12585},
year={2021}
}