Evaluating the Impact of Word Embeddings on Similarity Scoring in Practical Information Retrieval