Random Confirmation Mechanism for Speech Recognition Confidence Assessment

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional speech systems with deterministic confirmation mechanisms struggle to assess performance due to lack of confirmation on speech recognition results within certain confidence score ranges, leading to potential errors and inefficiencies in routing calls between automated responses and human operators.

Innovation Solution

A configurable confirmation mechanism that allows for random confirmation decisions based on configurable criteria, including a random confirmation range and integration with deterministic confirmation decisions, enabling more flexible and adaptive confirmation strategies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a deterministic confirmation mechanism is used with a pre-determined confirmation range of scores, then the confirmation process is simple and automated, but the system cannot assess performance on speech recognition results within uncertain confidence score ranges

Engineering Contradiction:

Improveconfirmation process simplicityVSAvoidperformance assessment capability

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The confirmation mechanism transitions from a static deterministic approach to a dynamic approach that can randomly select confirmation targets based on configurable criteria. This allows the system to adaptively choose which speech recognition results to confirm, enabling performance assessment across different confidence score ranges while maintaining automated operation through configurable randomness parameters.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system introduces randomization as a new parameter in the confirmation decision process, moving beyond fixed deterministic thresholds. By incorporating random selection with configurable criteria, the system can assess performance on previously unmeasured confidence score ranges while maintaining operational simplicity through parameter-based control.

Inventive Principle:

Principle #35Parameter changes

2Reliability

If deterministic confirmation is applied to all speech recognition results within a confidence range, then confirmation coverage is maximized, but call routing errors increase due to lack of confirmation on uncertain results

Engineering Contradiction:

Improveconfirmation coverageVSAvoidcall routing accuracy

Core Design Contradiction:

ReliabilityVSManufacturing precision

Solution Approach 1:

Instead of confirming all speech recognition results within a confidence range (excessive action), the system performs partial confirmation by randomly selecting a subset of results for confirmation. This partial action approach maintains sufficient confirmation coverage to improve routing accuracy while avoiding the errors caused by over-confident automated routing decisions.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The system implements feedback mechanisms where confirmed speech recognition results provide information about actual user intent, which can be used to refine future confirmation and routing decisions. This feedback loop enables the system to learn from confirmed interactions and improve call routing accuracy over time.

Inventive Principle:

Principle #23Feedback

3Productivity

If no confirmation is performed on speech recognition results with high confidence scores, then system efficiency is maintained, but errors in routing calls to human operators increase

Engineering Contradiction:

Improvesystem efficiencyVSAvoidrouting decision accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system performs partial confirmation by randomly selecting a subset of speech recognition results for confirmation, including some with high confidence scores. This partial action approach maintains system efficiency by not confirming every result while still providing sufficient confirmation coverage to improve routing decision accuracy through statistical sampling.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The system uses its own randomization mechanism to self-select which results require confirmation, eliminating the need for external intervention in the confirmation decision process. This self-service approach maintains efficiency while improving routing accuracy through automated, configurable random selection.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS7788103B2Random confirmation in speech based systems

Publication Date: 2010.08.31 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

Method and system are provided for performing random confirmation in a speech system. When a speech recognition result is received with an associated confidence score indicating a level of confidence with respect to the speech recognition result, a confirmation decision is made in terms of whether a confirmation is to be carried out based on the confidence score. The confirmation decision may be made in a random confirmation mode. A confirmation may be performed when the confirmation decision is to carry out a confirmation on the speech recognition result.