IMPROVING NEURAL LANGUAGE GENERATION WITH SPECTRUM CONTROL

Keyphrases

• Spectrum Control 100%
• Neural Text Generation 100%
• Transformer Model 66%
• Transformer-XL 66%
• Natural Language Processing 33%
• Control Method 33%
• Neural Model 33%
• Control Approach 33%
• Output Layer 33%
• Language Model 33%
• Vector Space 33%
• Benchmark Dataset 33%
• Prior Distribution 33%
• Machine Translation 33%
• Processing Task 33%
• Isotropy 33%
• Singular Values 33%
• Reparameterization 33%
• Word Representation 33%
• Embedding Matrix 33%
• Modeling Power 33%
• Contextual Embeddings 33%
• Singular Value Distribution 33%
• Output Embeddings 33%

Computer Science

• Generation Language 100%
• Optical Spectrum 100%
• Singular Value 50%
• Natural Language Processing 25%
• Machine Translation 25%
• Language Modeling 25%
• Control Method 25%
• Bidirectional Encoder Representations From Transformers 25%
• Processing Task 25%
• Embedding Matrix 25%
• Reparameterization 25%
• Spatial Distribution 25%