Using Deep-Q Network to Select Candidates from N-best Speech Recognition Hypotheses for Enhancing Dialogue State Tracking

Keyphrases

• Speech Recognition 100%
• Dialogue State Tracking 100%
• Deep Q-network 100%
• N-best 100%
• Automatic Speech Recognition 80%
• Tracking Accuracy 40%
• Utterance 40%
• Speech Recognition System 20%
• Top 10 20%
• Dialogue Systems 20%
• Noisy Environment 20%
• Stall 20%
• Accuracy Improvement 20%
• Tracking Method 20%
• Realistic Scenario 20%
• Improvement Factor 20%
• Deep Reinforcement Learning (deep RL) 20%
• Word Error Rate 20%
• State-based 20%
• Dialog State 20%

Computer Science

• Speech Recognition 100%
• Tracking State 100%
• Deep Q-Network 100%
• Recognition Result 60%
• Speech Recognition System 20%
• Noisy Environment 20%
• Tracking Method 20%
• Deep Reinforcement Learning 20%
• Word Error Rate 20%
• Dialog System 20%

Engineering

• Automatic Speech Recognition 100%
• Noisy Environment 20%
• Error Rate 20%
• Reinforcement Learning 20%