Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper redefines machine interpreting (MI) as a distinct subfield of speech translation, highlighting the disparity between high benchmark performance and poor user experience, termed the "accuracy illusion." By drawing insights from interpreting studies, the authors identify three critical design priorities—agency, grounding, and experience—that are essential for improving the usability of MI systems. The proposed framework aims to enhance real-time multilingual communication by addressing overlooked aspects of professional interpreting practice that current systems fail to support effectively.
The "accuracy illusion" reveals that even high-performing machine interpreting systems can falter in real-world communication, underscoring the need for a user-centered design approach.
Machine interpreting (MI), the live, real-time branch of speech translation, has achieved remarkable progress on standard benchmarks, with some systems approaching human parity on textual fidelity. Yet the user experience remains far inferior to interpreter-mediated communication, revealing what we term the \emph{accuracy illusion}: systems that appear accurate on paper but fail in practice to support smooth, goal-oriented interaction. This paper defines MI as a distinct subfield of speech translation, with its own characteristics and the need for evaluation methods grounded in communicative effectiveness rather than isolated fidelity metrics. Drawing on insights from interpreting studies, we identify critical dimensions of professional interpreting practice that are overlooked by current systems, and consolidate them into three interdependent design priorities for future MI: \emph{agency} (context-sensitive initiative and repair), \emph{grounding} (multimodal and discourse-level situational awareness), and \emph{experience} (adaptive improvement through real interaction). Together, these priorities chart a path toward closing the usability gap and enabling systems that can sustain authentic multilingual communication in real time.