How relevant is part-of-speech information to compute similarity between Greek verses in a graph database?