RPTCS: A Reinforced Persona-aware Topic-guiding Conversational System

• DOI: 10.18653/v1/2023.eacl-main.253
• Publication Date: 2023-09-09T20:54:31Z
• Authors: Zishan Ahmad; Kshitij Mishra; Asif Ekbal; Pushpak Bhattacha...
• Keywords: Persona; Casual; Dialog system

Although there has been a plethora of work on open-domain conversational systems, most the systems lack mechanism controlling concept transitions in dialogue. For activities like switching from casual chit-chat to task-oriented conversation, an agent with ability manage flow concepts conversation might be helpful. The user would find dialogue more engaging and receptive such if these were made while taking into account user’s persona. Focusing persona-aware transitions, we propose Reinforced Persona-aware Topic-guiding Conversational System (RPTCS). Due topic transition dataset, novel dataset creation which leads discourse drift set target depending persona speaker context conversation. To avoid scarcely available expensive human resource, entire data-creation process is mostly automatic human-in-loop only for quality checks. This created named PTCD used develop RPTCS two steps. First, maximum likelihood estimation loss-based model trained PTCD. Then this fine-tuned Reinforcement Learning (RL) framework by employing reward functions assure persona, topic, consistency non-repetitiveness generated responses. Our experimental results demonstrate strength proposed system respect strong baselines.