UniRaG: Unification, Retrieval, and Generation for Multimodal Question Answering With Pre-Trained Language Models

Keyphrases

• Context Retrieval 33%
• Contextual Diversity 16%
• Cross-encoder 16%
• Distractor 16%
• Diversity Training 16%
• Flan 16%
• Image Captioning 16%
• Multimodal Comprehension 16%
• Multimodal Question Answering 100%
• Negative Context 16%
• Pre-trained Language Model 100%
• Question Generation 50%
• Retrieval Strategies 16%
• Sequence Classification 16%
• Text Representation 16%
• Text-based Approach 33%
• Three-stage Framework 16%
• Top-k Retrieval 16%

Computer Science

• Diversity Training 33%
• Image Captioning 33%
• Pre-Trained Language Models 100%
• Sequence Classification 33%
• Textual Representation 33%
• Transformation Process 33%

Earth and Planetary Sciences

• Natural Language (Computers) 50%