Deep learning architectures to embed multi-word units into a vector space maximizing similarity between units of different sizes. - View it on GitHub