• Stars
    star
    173
  • Rank 220,124 (Top 5 %)
  • Language
    Python
  • Created over 3 years ago
  • Updated almost 3 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

NLP句子编码、句子embedding、语义相似度:BERT_avg、BERT_whitening、SBERT、SmiCSE

项目说明

本项目是句子embedding、语义匹配任务的代码复现,是在sbert库的基础上做的二次开发
包括以下方法:

  • BERT_avg
  • BERT_whitening
  • SBERT
  • SimCSE
  • ConSERT
  • ESimCSE

部分实验效果如下:

Chines-STS-B dev Chines-STS-B test
BERT-avg(CLS-token) 54.39 48.34
BERT-whitening 74.70 66.98
SBERT(SNLI) 74.40 72.41
SBERT(STS-B) 82.10 77.69
SimCSE 76.55 71.91
ConSERT 78.35 72.48
ESimCSE 78.82 71.81

更具体方法原理和效果,请看我的知乎博客https://zhuanlan.zhihu.com/p/450993543?

环境

  • python=3.6
  • torch=1.7
  • transformers=4.5.0

数据集

https://github.com/pluto-junzeng/CNSD 中的Chinese-SNLI数据集和Chinese-STS-B数据集