知识卡片
基准成绩受任务与评价条件限制
孤立短句的讽刺分类与完整对话中的语境理解是不同任务。比较人和模型时需相同样本、信息和评分规则,不能把一个数据集的68%或81%扩展成普遍超过人类;知识抽取、情感分类与事实正确也有不同指标。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- · 查看原书依据
The techniques you will learn, however, are powerful enough to create machines that can surpass humans in both accuracy and speed for some surprisingly subtle tasks. For example, you might not have guessed that recognizing sarcasm in an isolated Twitter message can be done more accurately by a machine than by a human. Well-trained human judges could not match the performance (68% accuracy) of a
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第01章《会读写的机器与自然语言处理》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-1.html)
- 结论依据:摘录中关于基准成绩受任务与评价条件限制的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:The techniques you will learn, however, are powerful enough to create machines that can surpass humans in both accuracy and speed for some surprisingly subtle tasks. For example, you might not have guessed that recognizing sarcasm in an isolated Twitter message can be done more accurately by a machine than by a human. Well-trained human judges could not match the performance (68% accuracy) of a