Enhancing Conference Interpreting with Computer-assisted Interpreting Tools: A Multi-agent System

The integration of artificial intelligence (AI) into computer-assisted interpreting (CAI) tools has the potential to revolutionize the field of simultaneous interpreting. However, current AI-enhanced CAI tools often struggle to fully address the real-world needs and challenges faced by professional interpreters. This study investigates the experiences and perceptions of professional conference interpreters regarding the use of CAI tools, following a targeted training programme. Participants include interpreters with experience working in public sector settings in Hong Kong. Through pre- and post-training surveys, we identify key limitations in existing tools and propose a novel, modular multi-agent CAI platform designed to overcome these challenges. In response, we present a conceptual prototype of a modular, multi-agent CAI platform designed to provide interpreters with adaptive, context-sensitive support. The system integrates three AI-driven agents: (1) Automatic Speech Recognition and Real-Time Translation, (2) Dynamic Terminology Management, and (3) Custom Domain Support. These agents work in parallel to provide interpreters with adaptive, context-specific support, reducing cognitive load and enhancing real-time decision-making. The system's architecture is grounded in the empirical needs of interpreters, emphasizing flexibility, customization, and seamless integration into existing workflows. This study presents the first practitioner-informed blueprint for an AI-driven, multi-agent CAI platform, bridging the gap between interpreter expertise and technological innovation. We discuss the implications of this system for professional practice, interpreter training, and future research directions, including the need for empirical validation to assess the platform's impact on interpreter cognition and performance. By proposing a CAI tool that adapts to the diverse needs of interpreters across various settings, this work represents a significant step towards the development of truly interpreter-centred AI support systems.