知识卡片

普通读书笔记卡 · 6776.a · 生成模型:GPT-6

代词先行与后指都会造成歧义

理解等级:中级

提及可能在明确名称前出现,也可能有多个候选。Winograd式最小变式可检查系统是否利用关系和语境,而非只选最近名词;困难样例揭示能力边界,不能由此断言某整类模型绝不可能处理共指。

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

      As a human, reading this text, you understand that Gebru, she, and her all relate. But this bit of commonsense reasoning and coreference resolution is a lot trickier for a machine. Automatic coreference resolution can be especially difficult for a machine if she is mentioned before Gebru (a linguistic pattern called cataphora). Nonetheless, the transformer-based Coreferee language model can handle even this challenging kind of

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第11章《信息抽取与知识图谱》(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-11.html) - 结论依据:摘录中关于代词先行与后指都会造成歧义的定义、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:As a human, reading this text, you understand that Gebru, she, and her all relate. But this bit of commonsense reasoning and coreference resolution is a lot trickier for a machine. Automatic coreference resolution can be especially difficult for a machine if she is mentioned before Gebru (a linguistic pattern called cataphora). Nonetheless, the transformer-based Coreferee language model can handle even this challenging kind of