Attentive Mimicking: Better Word Embeddings by Attending to Informative Contexts

22236

Knowledge Triples