Abstract
Proposed herein are techniques that can facilitate transforming today's "tone-deaf" and static artificial intelligence (AI) receptionists into a self-improving, intelligent system purpose-built for seamless and accurate caller routing in the enterprise. The proposed system introduces a dynamic conversational engine that uses just-in-time (JIT) vocabulary injection for real-time accuracy, an automated learning pipeline that learns from every human-led correction, and a multi-modal fusion engine that intelligently weighs phonetic, acoustic, and contextual signals. This approach empowers enterprises to eliminate caller frustration, drastically improve name recognition accuracy, and bridge the dangerous gap between a caller's intent and an automated system's ability to understand the caller's intent.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Shrivastava, Puneet; Dhar, Rajarshee; Mohan R, Ram; and Lukacsy, Gergely, "A SELF-IMPROVING AI SYSTEM FOR PHONETIC NAME RESOLUTION VIA A CLOSED-LOOP, MULTI-MODAL, AUTOMATED LEARNING ARCHITECTURE", Technical Disclosure Commons, (May 13, 2026)
https://www.tdcommons.org/dpubs_series/10095