Multi-Interest Refinement by Collaborative Attributes Modeling for Click-Through Rate Prediction

Keyphrases

• Transformer 100%
• Click-through 100%
• Multi-interest 100%
• Interest Representation 100%
• Attribute Modeling 100%
• Rate Prediction 100%
• Transformer-based 50%
• User Interest 50%
• Benchmark Dataset 50%
• Learning Interest 50%
• Real-world Benchmark 50%
• Prediction Task 50%
• Attention Weight 50%
• Routing Module 50%
• Item-specific 50%
• Diversity Loss 50%
• Attribute Relevance 50%
• Industrial Datasets 50%

Computer Science

• Interest Representation 100%
• Routing Module 50%
• Learning Interest 50%
• Probability Mass 50%
• Assign Probability 50%