基于论文ESimCSE进行复现,基于STS-B训练集 + 额外数据 进行训练,在中文STS-B的验证集spermanr相关性得分为0.7201. | |
论文参考: | |
@inproceedings{Wu2021ESimCSEES, | |
title={ESimCSE: Enhanced Sample Building Method for Contrastive Learning of Unsupervised Sentence Embedding}, | |
author={Xing Wu and Chaochen Gao and Liangjun Zang and Jizhong Han and Zhongyuan Wang and Songlin Hu}, | |
booktitle={International Conference on Computational Linguistics}, | |
year={2021} | |
} |