Mixed-Mode Dictation Correction With Context-Aware Voice Interpretation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current voice input solutions struggle to accurately distinguish between textual content and voice commands, leading to misinterpretations that diminish the user experience.

Innovation Solution

A data processing system that employs a multistage analysis using natural language processing and machine learning models to interpret voice inputs, incorporating context information and user feedback to improve the distinction between textual content and commands, with mechanisms for corrective actions and learning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If voice input is interpreted using basic speech recognition, then processing speed is maintained, but interpretation accuracy between textual content and commands deteriorates

Engineering Contradiction:
Improveinterpretation accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The voice input interpretation is divided into multiple stages: initial speech-to-text conversion, followed by separate analysis of command indicators versus textual content. The system segments the interpretation process into distinct functional components that can be evaluated independently, improving accuracy without proportionally increasing overall complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system introduces an intermediary analysis layer that sits between basic speech recognition and final interpretation. This intermediary stage identifies command indicators and contextual cues, mediating between raw speech input and the final determination of whether the input is a command or textual content, thereby improving accuracy.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If contextual analysis is added to voice interpretation, then distinction accuracy between text and commands improves, but processing time increases

Engineering Contradiction:
Improvedistinction accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary identification of potential command indicators during the initial speech-to-text conversion phase. By pre-identifying key command-related keywords and structures before full contextual analysis, the system reduces the computational burden of subsequent processing while maintaining high distinction accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system applies contextual analysis selectively rather than uniformly to all voice inputs. When command indicators are detected, full contextual analysis is applied; for clear textual inputs, minimal processing is performed. This partial application of analysis reduces average processing time while maintaining high accuracy for ambiguous cases.

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If machine learning models are trained with more data, then interpretation reliability improves, but model complexity and training resources increase

Engineering Contradiction:
Improveinterpretation reliabilityVSAvoidmodel complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The machine learning model is segmented into multiple specialized components: one model identifies command indicators, another distinguishes command types, and a third handles textual content classification. This segmentation allows each sub-model to be simpler and more specialized, improving overall reliability without requiring a single monolithic complex model.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system employs a multi-functional interpretation framework where a single processing architecture handles both command identification and textual content classification. This universal framework reuses the same base speech-to-text engine and contextual analysis mechanisms for both functions, reducing overall system complexity while improving reliability through consistent processing.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4189672B1Compounding corrective actions and learning in mixed mode dictation
Publication Date: 2026.03.11 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP4189672B1 patent drawingFigure 1A
  • EP4189672B1 patent drawingFigure 1B
  • EP4189672B1 patent drawingFigure 2

AI summary

Techniques performed by a data processing system for processing voice content received from a user herein include receiving a first audio input from the user comprising a mixed-mode dictation, analyzing, using one or more machine learning (ML) models, the first audio input to obtain a first interpretation of the mixed-mode dictation, presenting the first interpretation to the user in an application on the data processing system, receiving a second audio input from the user comprising a corrective command, analyzing the second audio input to obtain a second interpretation of the restatement of the mixed-mode dictation presenting the second interpretation to the user, receiving an indication from the user that the second interpretation is a correct interpretation of the mixed-mode dictation, and modifying the operating parameters of the one or more machine learning models to interpret the subsequent instances of the mixed-mode dictation based on the second interpretation.