This library delivers numerical representations (features) for short texts or sentences, which can be used as input to any machine learning task later on. Think of it as an unsupervised version of FastText, and an extension of word2vec (CBOW) to sentences.