Creating sentiment polarity lexicons is labor intensive. Automatically translating them from resourceful languages requires in-domain machine translation systems, which rely on large quantities of bi-texts. In this paper, we propose to replace machine translation by transferring words from the lexicon through word embeddings aligned across languages with a simple linear transform. The approach leads to no degradation, compared to machine translation, when tested on sentiment polarity classification on tweets from four languages.
@article{arxiv.1612.05202,
title = {Building a robust sentiment lexicon with (almost) no resource},
author = {Mickael Rouvier and Benoit Favre},
journal= {arXiv preprint arXiv:1612.05202},
year = {2016}
}