Victor Prokhorov, Mohammad Taher Pilehvar, Dimitri Kartsaklis, Pietro Lió, Nigel Collier
TL;DR本文提出一种方法学习语言的词汇，并使用图嵌入技术和跨语言向量空间映射方法结合本体源和语料库来提高现有词向量的覆盖范围以及加强其词汇知识。所提出的方法在Rare Word Similarity dataset上表现出10%的性能提升。
Abstract
We propose a methodology that adapts graph embedding techniques (DeepWalk (Perozzi et al., 2014) and node2vec (Grover and Leskovec, 2016)) as well as cross-lingual vector space mapping approaches (Least Squares a