Natural Language Processing(7)

• Efficient human-like semantic representations via the Information Bottleneck principle
• Specializing Word Embeddings (for Parsing) by Information Bottleneck
• Specializing Word Embeddings (for Parsing) by Information Bottleneck
• How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings
• Correlations between Word Vector Sets
• Revealing the Dark Secrets of BERT
• nlp-category-test