Answering Machine Detection for Reassigned Number Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contact centers face challenges in determining whether an outbound call is reaching the intended individual, as telephone numbers may become obsolete due to reassignment, leading to potential violations of regulations like the TCPA and increased costs due to incorrect identification of answering machines or live persons.
Innovation Solution
The implementation of a system that analyzes the audio and transcript of answering machine greetings to determine if a call is directed to a reassigned number by using timer-based and linguistic approaches, including the use of Real-Time Speech Analytics (RTSA) to differentiate between live person and machine responses, thereby providing accurate classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional answering machine detection methods are used, then the system can identify automated responses, but it cannot reliably detect reassigned numbers leading to regulatory violations
Solution Approach 1:
The patent segments the greeting analysis into multiple independent detection dimensions: timer-based detection (measuring greeting duration and pause intervals), linguistic analysis (identifying machine-specific phrases and patterns), and audio feature analysis (detecting synthetic voice characteristics). Each dimension operates independently and contributes to the overall reassigned number detection, allowing the system to achieve high reliability through multiple verification paths rather than relying on a single complex method.
Solution Approach 2:
The patent introduces an intermediary comparison mechanism that matches current call greetings against a database of known reassigned number greetings and templates. This intermediary layer enables the system to detect reassigned numbers by comparing patterns without requiring direct knowledge of number reassignment status, thereby improving detection accuracy while maintaining manageable system complexity through reusable comparison templates.
2Measurement precision
If multiple detection methods are implemented to improve accuracy, then reassigned number detection improves, but processing time and computational resources increase
Solution Approach 1:
The patent implements a tiered detection approach where timer-based detection (a simple, fast method) is applied first to all calls. Only calls that exhibit characteristics suggestive of reassigned numbers proceed to more computationally intensive linguistic and audio analyses. This partial application of advanced detection methods to only relevant cases maintains high detection precision while minimizing overall processing time and resource consumption.
Solution Approach 2:
The system performs preliminary timer-based detection and basic pattern matching before committing to full linguistic and audio analysis. By pre-filtering calls using quick, low-computation methods, the system prepares data in advance and avoids unnecessary processing of calls that clearly do not involve reassigned numbers, thereby reducing total processing time while maintaining detection precision for problematic cases.
3Reliability
If enhanced AMD techniques are used to detect reassigned numbers, then regulatory compliance improves, but implementation complexity and cost increase
Solution Approach 1:
The patent designs the enhanced AMD system with universal components that serve multiple functions: the timer-based detection module not only identifies greeting duration patterns but also provides basic call routing decisions; the linguistic analysis engine both detects reassigned numbers and generates compliance reporting data; the audio feature analysis serves both detection and quality assurance purposes. This multi-functionality reduces implementation complexity by reusing the same components across different objectives rather than building separate systems for each function.
Solution Approach 2:
The system incorporates self-learning capabilities where detection results from initial calls automatically update the database of known reassigned number patterns. As the system processes more calls, it automatically refines its detection models and expands its pattern library without requiring manual configuration or external intervention. This self-service approach reduces implementation burden and makes the system progressively easier to operate as it accumulates experience.
Data Source
AI summary
Answering machine detection (“AMD”) processes in a contact center is used to derive and store AMD meta-data about a known greeting from an automatic voice messaging capability (“AVMC”) on a telephone call to a known number. The AMD meta-data is used in subsequent calls to that known number wherein the greeting detected is analyzed using the AMD meta-data to make a comparison if the current greeting is the same of that previously encountered. If the current greeting is the same as previously encountered, this suggests the number has not been reassigned. If different, this suggests the number was reassigned. If the distinction is not clear, then the agent may perform a manual comparison of the audio speech and/or a transaction of the greeting. If the number is reassigned, then a number validation for the intended individual can be performed to verify whether the number was reassigned.