Applying Exponential Family Embeddings in Natural Language Processing to Analyze Text