第8章:在8.2.5节的语义记忆实现中,系统会自动提取实体和关系构建知识图谱。请分析:这种自动提取的准确性如何?在什么情况下可能会提取出错误的实体或关系?如何设计一个"知识图谱质量评估"机制? #148
Unanswered
LiuYuPeng1101
asked this question in
💬 Exercises & Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
准确率/召回率:人工抽样校对实体类型、属性、关系,并计算 F1
置信度统计:记录模型自带的置信度或对数概率,监控低置信实体的占比。
最准确的是第2种,根据生产数据进行反馈
Beta Was this translation helpful? Give feedback.
All reactions