ASR Proxy Routing for Speech Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current interactive response systems, such as IVR, face challenges in providing a satisfactory customer experience due to limitations in automated speech recognition (ASR) systems, which often require users to speak in a constrained manner and struggle with accents, dialects, and background noise, leading to high error rates and frustration.
Innovation Solution
An interactive response system that combines Automated Speech Recognition (ASR) and Human Speech Recognition (HSR) subsystems, using an ASR proxy to route utterances to either ASR or HSR based on confidence prediction, historical results, and user-specific voice recognition, allowing for real-time adjustments in resource allocation and grammar tuning to improve recognition accuracy and user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If ASR systems are used to automate customer service interactions, then productivity and cost efficiency are improved, but recognition accuracy and user satisfaction deteriorate due to constraints on speech patterns and difficulty with accents and background noise
Solution Approach 1:
The patent introduces an intermediary component that bridges ASR and HSR systems. This intermediary manages the routing of customer interactions, determining when to use automated speech recognition and when to escalate to human speech recognition based on confidence scores and interaction context, thereby maintaining both automation efficiency and recognition reliability
2Measurement precision
If ASR systems require users to speak in constrained patterns, then recognition accuracy is improved, but ease of operation deteriorates as users must adapt to computer-like speech patterns
Solution Approach 1:
The system dynamically adjusts the level of constraint applied to user speech. Rather than enforcing fixed grammatical constraints, the system adapts its recognition patterns based on the interaction context, confidence scores, and user behavior, allowing users to speak more naturally while maintaining accuracy through flexible pattern matching and contextual understanding
3Reliability
If HSR resources are used for all customer interactions, then recognition accuracy is improved, but productivity and cost efficiency deteriorate due to high agent turnover and training requirements
Solution Approach 1:
The patent segments customer service interactions into different handling paths: simple, high-confidence interactions are routed to ASR for automated handling, while complex or low-confidence interactions are routed to HSR. This segmentation allows the system to leverage the efficiency of automation for suitable cases while reserving human resources for situations requiring their expertise, thereby improving overall productivity without sacrificing accuracy
4Productivity
If ASR systems are used exclusively, then cost efficiency is improved, but adaptability deteriorates due to inability to handle diverse speech patterns, accents, and dialects
Solution Approach 1:
The system creates a universal customer service platform that can handle multiple speech patterns, accents, and dialects by combining ASR and HSR capabilities. The hybrid architecture allows the system to function effectively across diverse linguistic scenarios, adapting to different user groups while maintaining cost efficiency through selective automation
Data Source
AI summary
An interactive response system mixes HSR subsystems with ASR subsystems to facilitate overall capability of user interfaces. The system permits imperfect ASR subsystems to nonetheless relieve burden on HSR subsystems. An ASR proxy is used to implement an IVR system, and the proxy dynamically selects one or more recognizers from a language model and a human agent to recognize user input. Selection of the one or more recognizers is based on factors such as confidence thresholds of the ASRs and availability of human resources for HSRs.