Calling word embeddings "models" is a bit misleading.

[xuf12]

Thanks for putting this together! However, the embedding vectors are just the weights from a shallow NN, which offers far less info than a complete language model (just like the weights from the first layer of training ResNet and ResNet itself). I think we may want to change the title so it doesn't mislead people into believing a word vector is a model...