← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6647 · 生成模型:GPT-6

抽取式与生成式产出有不同依据

理解等级:中级

抽取式方法从已有材料选择内容,生成式方法组合出新的文本。前者仍可能取错或丢失上下文,后者也未必保证新颖、真实或无抄录。用于代码时,语法看似完整也还需要核对意图、依赖和实际行为。

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

      And these tools just got a whole lot smarter. Older code generation tools were extractive. Extractive text generation algorithms find the most relevant text in your history and simply regurgitate it verbatim, as a suggestion. So if the phrase prosocial artificial intelligence appears a lot in the text an algorithm was trained on, autocomplete will recommend the term artificial intelligence to follow prosocial rather than

    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:摘录中关于抽取式与生成式产出有不同依据的定义、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:And these tools just got a whole lot smarter. Older code generation tools were extractive. Extractive text generation algorithms find the most relevant text in your history and simply regurgitate it verbatim, as a suggestion. So if the phrase prosocial artificial intelligence appears a lot in the text an algorithm was trained on, autocomplete will recommend the term artificial intelligence to follow prosocial rather than