Learning Call Analysis System for Contact Center Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated dialer systems in contact centers face inefficiencies due to incorrect detection of pre-recorded audio as live speakers, leading to agents mistakenly interacting with machines, which results in false positives and reduced productivity, and existing solutions either disable Answering Machine Detection (AMD) or require frequent relearning of audio recordings.
Innovation Solution
Implementing a learning call analysis system that generates and compares audio fingerprints to identify pre-recorded audio, allowing automated dialers to differentiate between live speakers and recordings, thereby optimizing the routing of communications and reducing false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Answering Machine Detection (AMD) is enabled in automated dialer systems, then pre-recorded audio can be identified and routed away from agents, but false positives occur where live speakers are incorrectly classified as recordings
Solution Approach 1:
The system implements feedback by analyzing calls that were routed to agents and determining whether they actually contained pre-recorded audio. This feedback loop allows the system to learn from previous routing decisions and improve future detection accuracy, reducing false positives while maintaining high detection rates
Solution Approach 2:
The system performs preliminary analysis by generating audio fingerprints and comparing them against a database of known recording fingerprints before routing calls to agents. This preliminary action enables the system to proactively identify pre-recorded audio and prevent false routing decisions before they occur
2Productivity
If AMD is disabled to avoid false positives, then live speakers are not mistakenly routed away from agents, but pre-recorded audio is not identified and agents must handle all calls manually
Solution Approach 1:
The system enables self-service by automatically analyzing and classifying audio recordings without requiring manual agent intervention. The automated dialer system independently determines whether calls contain pre-recorded audio and routes them appropriately, freeing agents to focus on live speaker interactions while maintaining high detection accuracy
3Reliability
If traditional audio analysis methods are used to detect pre-recorded audio, then the system can identify recordings, but the system requires frequent relearning and cannot adapt to new or modified recordings
Solution Approach 1:
The system implements dynamics by continuously learning and adapting to new pre-recorded audio formats. The feedback mechanism allows the system to update its detection models based on newly encountered recordings, enabling it to adapt to changes in recording formats, languages, and styles without requiring complete relearning
Solution Approach 2:
The system performs preliminary fingerprint generation and database creation to establish a baseline for detecting pre-recorded audio. This preliminary action creates a structured framework that can efficiently compare new audio against known patterns, reducing the need for frequent comprehensive relearning while maintaining detection accuracy
4Reliability
If automated dialers route all calls to agents without analysis, then no false positives occur, but contact center efficiency is reduced due to agents handling pre-recorded audio
Solution Approach 1:
The system replaces manual agent analysis with automated audio fingerprinting and comparison technology. This substitution enables rapid, accurate identification of pre-recorded audio without requiring agent time and effort, significantly improving contact center efficiency while maintaining high routing accuracy through automated decision-making
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A system and method are presented for learning call analysis. Audio fingerprinting may be employed to identify audio recordings that answer communications. In one embodiment, the system may generate a fingerprint of a candidate audio stream and compare it against known fingerprints within a database. The system may also search for a speech-like signal to determine if the end point contains a known audio recording. If a known audio recording is not encountered, a fingerprint may be computed for the contact and the communication routed to a human for handling. An indication may be made as to if the call is indeed an audio recording. The associated information may be saved and used for future identification purposes.