Skip to content

Latest commit

 

History

16 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 

Repository files navigation

aliwentian

很多品牌词比较重要, 品牌匹配 v1: 传统匹配 v2: 基于bert

  1. query: char embedding + word embedding + mean_pooling + dense doc: word embedding + mean_pooling all data mrr@10: 0.039

  2. query: char embedding + word embedding + mean_pooling + dense doc: word embedding + attention all data mrr@10: 0.044

  3. query: char embedding + word embedding + attention + dense doc: word embedding + attention all data mrr@10: 0.050 线上: 0.07

  4. query: char embedding + word embedding + block(conv*3 + attention) doc: word embedding + block(conv + attention) 不如3

  5. query、doc都只是用word embedding 简化block结构(conv*1 + attention) 繁体转简体,全角转半角

  6. 传统模型分数到过mrr@10: 0.147

  7. 类似lebert, 在0层token embedding 融合 word embedding, 收敛上是更快的, 有没有没有对比验证, 单估计不会导致降低

  8. 没有尝试过预训练,最后分数0.251,a榜最高0.42, 分数不理想,不过学习到了一些知识吧

  9. 所有层cls向量简单加权融合, 权重为可训练的参数 效果变差

  10. 训练好后, 加入hard sample, 微调最后几层,有提升

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages