Semantics- and Syntax-related Subvectors in the Skip-gram Embeddings
We show that the skip-gram embedding of any word can be decomposed into two subvectors which roughly correspond to semantic and syntactic roles of the word.
READ FULL TEXTWe show that the skip-gram embedding of any word can be decomposed into two subvectors which roughly correspond to semantic and syntactic roles of the word.
READ FULL TEXTUse your Google Account to sign in to DeepAI