Word embeddings for the Polish language

Abstract

We present a dataset of word embeddings for the Polish language. Presented embeddings can be used as an input for Artificial Intelligence methods as an alternative for one-hot representation. Spatial relations between embeddings reflect relations such as alternatives and analogies. This improves generalization of methods using presented embeddings. Data from Wikipedia has been used together with skip-gram and contitous-bag-of-words methods introduced originally for English language by Mikolov et al. Current version of embeddings can be downloaded from http://publications.ics.p.lodz.pl/2016/word embeddings/.