Relations such as Hypernymy: Identifying and Exploiting Hearst Patterns in Distributional Vectors for Lexical Entailment. (arXiv:1605.05433v1 [cs.CL])

We consider the task of predicting lexical entailment using distributional vectors. We focus experiments on one previous classifier which was shown to only learn to detect prototypicality of a word pair. Analysis shows that the model single-mindedly learns to detect Hearst Patterns, which are well known to be predictive of lexical relations. We present a new model which exploits this Hearst Detector functionality, matching or outperforming prior work on multiple data sets.

from cs.AI updates on arXiv.org http://ift.tt/1U0ca4A
via IFTTT