1 paper
Seunghyun Yoon, Pablo Estrada, Kyomin Jung
We propose a novel approach to learn word embeddings based on an extended version of the distributional hypothesis. Our model derives word embedding vectors using the etymological…