Unsupervised Dense Information Retrieval with Contrastive Learning
Contriver
openreview, github
总体来说使用对比学习,正样本是啥呢,就是同文档内部的句子为正样本之类的。
Unsupervised Legal Evidence Retrieval via Contrastive Learning with Approximate Aggregated Positive
github
又是一整个无监督对比学习,在一个司法的场景下,标注了一堆新的数据