End-to-End Gujarati Task-Oriented Dialogue Management using Reinforcement Learning

  • Unique Paper ID: 163264
  • Volume: 10
  • Issue: 11
  • PageNo: 2904-2911
  • Abstract:
  • Nowadays, there's an increased demand for dialogue systems in local languages due to the ongoing need for continuous support in specific service domains. Ra-ther than relying solely on human resources, dialogue systems offer a viable solu-tion. Dialogue management plays a pivotal role in determining the most effective actions for the system at each stage. In this study, we introduce a task-oriented dialogue system for Gujarati language, leveraging reinforcement learning. This system comprises three key components: natural language understanding (NLU), Dialogue Management (DM), and Natural Language Generation (NLG). Our model seamlessly interacts with databases, extracting valuable information. Rein-forcement learning is employed specifically for the DM, employing an enhanced Deep Q-learning Network (DQN) strategy to bolster the agent's resilience against environmental noise. Additionally, we propose a unified model for the NLU module, demonstrating its effectiveness through experiments conducted on Guja-rati dialogue datasets. The results showcase the superior performance of our model over the conventional rule-based multi-turn dialogue system for Gujarati dialogues.

Cite This Article

  • ISSN: 2349-6002
  • Volume: 10
  • Issue: 11
  • PageNo: 2904-2911

End-to-End Gujarati Task-Oriented Dialogue Management using Reinforcement Learning

Related Articles