AI-Driven Voice Inputs for Speech Engine Testing in Conversational Systems
Vigneshwaran Jagadeesan Pugazhenthi, Gokul Pandy, Baskaran Jeyarajan, Aravindhan Murugan · 2025
Testing voice-based applications, such as conversational or traditional IVR (Interactive Voice Response) systems, relies heavily on speech to determine the caller's intent. Accurate recognition of this intent ensures the conversation progresses smoothly, allowing the system to retrieve the right information for better service. Each individual's voice has unique characteristics-such as accents, frequencies, speech styles, and paces-which can significantly vary across different callers. Therefore, ensuring that conversational IVRs are equipped with high-quality Automatic Speech Recognizers (ASRs) is crucial for processing these variations and accurately responding to user requests. This paper explores the role of Artificial Intelligence (AI) in enhancing Automatic Speech Recognizer (ASR) systems to recognize speech variations and how AI can also generate diverse speech inputs in different accents, tones, and paces for effective testing.