We present a new method for estimating vector space representations of words: embedding learning by concept induction. We test this method on a highly parallel corpus and learn semantic representations of words in 1259 different languages in a single common space. An extensive experimental evaluation on crosslingual word similarity and sentiment analysis indicates that concept-based multilingual embedding learning performs better than previous approaches.
@article{arxiv.1801.06807,
title = {Embedding Learning Through Multilingual Concept Induction},
author = {Philipp Dufter and Mengjie Zhao and Martin Schmitt and Alexander Fraser and Hinrich Schütze},
journal= {arXiv preprint arXiv:1801.06807},
year = {2018}
}