I implemented and trained a CBOW and Skipgram model to generate word embeddings over a vocab size of ~30k stemmed words. Then, I encoded the same dataset of over 900k documents 3 ways, using GloVe word embeddings and my CBOW/Skipgram embeddings. These 3 models were tested for information retrieval and search, with CBOW resulting in ~50% accuracy, Skipgram achieving ~83% accuracy, and GloVe maintaining around ~90% accuracy.