知识卡片

普通读书笔记卡 · 6664.e · 生成模型:GPT-6

跨域低分需与实际基准比较

理解等级:中级

产品评论约55.7%的结果提示迁移困难,但只有类别相近均衡时才接近随机硬币基准。应检查多数类、标签噪声、领域词及否定处理,再判断改进方向;加入n-gram可提供线索,却不保证自动消除全部否定问题。

这张卡片看不懂?

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

      0.557... #1 Correct predictions are when the predicted sentiment is the same as the label from the dataset. So your naive Bayes model does a poor job predicting whether the sentiment of a product review is positive (thumbs up) or negative (thumbs down), only a little better than a coin flip. One reason for this subpar performance is that your vocabulary from the casual_tokenize product

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第02章《思维的词元》(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-2.html) - 结论依据:摘录中关于跨域低分需与实际基准比较的定义、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:0.557... #1 Correct predictions are when the predicted sentiment is the same as the label from the dataset. So your naive Bayes model does a poor job predicting whether the sentiment of a product review is positive (thumbs up) or negative (thumbs down), only a little better than a coin flip. One reason for this subpar performance is that your vocabulary from the casual_tokenize product