Improving Negative Sampling for Word Representation using Self-embedded Features

Open in new window