---
license: mit
---
These direction vectors of antonyms can be used to calculate fasttext interpretable embeddings on the fly, solving the OOV problem of other interpretable embeddings. 
Simply calculate cosine similarity for each row.