Natural Language Understanding Wiki
Advertisement

Some seem to believe that linear combination can encode everything? Tensor (or matrix) factorization has been applied to word embeddings, selectional preferences and verb subcategoriation, what else...?