BriefGPT.xyz
May, 2023
使用下一句预测和互信息在潜空间中评估开放领域对话
Evaluating Open-Domain Dialogues in Latent Space with Next Sentence Prediction and Mutual Information
HTML
PDF
Kun Zhao, Bohao Yang, Chenghua Lin, Wenge Rong, Aline Villavicencio...
TL;DR
提出一种利用条件变分自编码器和下一句预测目标来处理开放领域对话的评估方法，该方法在潜空间中运用互信息建模以考虑文本语义相似性，实验证明这种学习基方法在处理含有语义不同的多个恰当回复时优于其他模型。
Abstract
The long-standing one-to-many issue of the
open-domain dialogues
poses significant challenges for
automatic evaluation
methods, i.e., there may be multiple suitable responses which differ in semantics for a given
→