文本生成模型，避免生成结果大多为训练数据里占比大的，有哪些方法？

lipiji / SongNet

Code for ACL 2020 paper "Rigid Formats Controlled Text Generation"：https://www.aclweb.org/anthology/2020.acl-main.68/

MIT License

230 stars 40 forks source link

Closed guotong1988 closed 3 years ago

guotong1988 commented 3 years ago

请问下大神 @lipiji 多谢多谢

guotong1988 commented 3 years ago

这个貌似是个https://github.com/hengyicai/ContrastiveLearning4Dialogue