知识卡片
去除字符差异会损失语言信息
ASCII化或去附加符号可合并部分拼写变体,也会消除原本有用的语言区别并使不同字串碰撞。需保留映射及原文,按任务决定是否采用;非ASCII文字不必都先转拉丁字母,归一化不保证对所有语言提高质量。
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
This multilingual dataset will give you a chance to learn how to deal with diacritics and other embellishments common in non-English words. To keep it interesting, you will remove these character embellishments and other giveaways in the Unicode characters of multilingual text. That way, your model can learn the patterns you really care about, rather than using these “cheats” to do things the easy way.
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第08章《循环网络与长短期记忆》(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-8.html)
- 结论依据:摘录中关于去除字符差异会损失语言信息的定义、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:This multilingual dataset will give you a chance to learn how to deal with diacritics and other embellishments common in non-English words. To keep it interesting, you will remove these character embellishments and other giveaways in the Unicode characters of multilingual text. That way, your model can learn the patterns you really care about, rather than using these “cheats” to do things the easy way.