lukecq1231 / kim

Knowledge-based Inference Model
Apache License 2.0
55 stars 13 forks source link

Lexical Semantic Relations Confusion #3

Open tc-yue opened 5 years ago

tc-yue commented 5 years ago

您的论文中在4.1中说同义词的识别[dog,wolf]=0,共享上位词 [dog,wolf]=1, 然后对于五维的关系特征利用公式4的indication function进行计算。可是例如词对是dog,wolf的时候,这种方式是无法识别entailment还是contradiction的。不知自己理解是否有误,期待您的回复。

lukecq1231 commented 5 years ago

论文中公式4主要是用来帮助词与词之间的对齐的。无论两个词是同义词或是共享上位词关系,它们都更有可能对齐。公式7和8中利用5维关系特征的方式则可以区分entailment和contradiction。