Deep Reinforcement Learning for natural language understanding and Dialogue Systems

Innovations in neural dialogue models promise to improve conversational AI. However, these models frequently lack foresight, handling answers one by one without considering long-term consequences. Traditional NLP techniques included reinforcement learning to address this. Deep reinforcement learning...

Full description

Saved in:
Bibliographic Details
Published in2023 6th International Conference on Recent Trends in Advance Computing (ICRTAC) pp. 736 - 741
Main Authors R, Prasanna Kumar, K, Venkatraman, B, Siva Jyothi Natha Reddy, Sree, V V V Bhagya, Anvitha, Vemireddy, K, Manojna, R, Sudarshan
Format Conference Proceeding
LanguageEnglish
Published IEEE 14.12.2023
Subjects
Brain modeling
chatbot dialogues
Chatbots
Coherence
coherent dialogues
Computational modeling
conversational agents
Deep reinforcement learning
foresight
neural dialogue generation
Oral communication
reinforcement learning
Technological innovation
virtual agents
Online AccessGet full text

Cover

More Information
Summary:Innovations in neural dialogue models promise to improve conversational AI. However, these models frequently lack foresight, handling answers one by one without considering long-term consequences. Traditional NLP techniques included reinforcement learning to address this. Deep reinforcement learning predicts future rewards in chatbot talks, according to this study. The model simulates agent interactions using policy gradient approaches, prioritizing sequences with essential conversational characteristics such as informativeness and coherence. Variety, answer length, and human assessment are emphasized in the evaluation of dialogue simulations, demonstrating the potential for interesting, protracted talks. This study is the first step toward developing a neural conversational model that prioritizes long-term discourse success.
DOI:10.1109/ICRTAC59277.2023.10480771