An approach more along the lines of machine learning would be to use what word2vec originated from at Berkeley Lab https://www.kaggle.com/c/word2vec-nlp-tutorial/forums/t/1234...