知识卡片
预处理与任务标签需要原文对照
毒性数据的六标签可同时成立,需要固定标签顺序和原列名。原数据常见severe_toxic与identity_hate,书表简写severe与hate若未显式重命名会读取失败;不要靠前五行推断全数据按无毒排序。
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
A toxic dataset You can download the “Toxic Comment Classification Challenge” dataset (archive.zip) from kaggle.com.35 You can put the data in your $HOME/.nlpia2-data/ directory with all the other large datasets from the book, if you like. When you unzip the archive.zip file, you’ll see it contains the training set (train.csv) and test set (test.csv) as separate CSV files. In the real world, you would probably
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第09章《可堆叠的Transformer》(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-9.html)
- 结论依据:摘录中关于预处理与任务标签需要原文对照的定义、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:A toxic dataset You can download the “Toxic Comment Classification Challenge” dataset (archive.zip) from kaggle.com.35 You can put the data in your $HOME/.nlpia2-data/ directory with all the other large datasets from the book, if you like. When you unzip the archive.zip file, you’ll see it contains the training set (train.csv) and test set (test.csv) as separate CSV files. In the real world, you would probably