Don’t count, predict! A systematic comparison of context-counting vs. context-predicting semantic vectors