Word2Vec: Optimal Hyper-Parameters and Their Impact on NLP Downstream Tasks