Context-Aware Speech Recognition Model Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing speech recognition systems often rely on users to select the correct speech recognition model, leading to inaccurate results due to incorrect model selection.

Innovation Solution

Implementing neural networks that incorporate multiple speech recognition models and utilize context information such as IP address, geographic location, and search history, along with time-independent audio characteristics, to automatically select and combine model outputs for improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If users manually select a speech recognition model from multiple available models, then the system can support multiple languages and dialects, but the accuracy deteriorates when users select an incorrect model

Engineering Contradiction:

Improvesupport for multiple languages and dialectsVSAvoidspeech recognition accuracy

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system automatically selects the appropriate speech recognition model by analyzing context information (such as geographic location, IP address, and search history) and audio characteristics, eliminating the need for manual user selection and preventing incorrect model choices

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system changes the selection criterion from manual user choice to automated analysis of context parameters and audio features, dynamically determining the most suitable speech recognition model based on current conditions

Inventive Principle:

Principle #35Parameter changes

2Reliability

If the system incorporates context information and multiple speech recognition models, then the accuracy improves, but the device complexity increases

Engineering Contradiction:

Improvespeech recognition accuracyVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the complex model selection task into distinct components: extracting context information (location, IP, search history), analyzing audio characteristics, and using a statistical classifier to select the appropriate model, making the overall system more manageable and efficient

Inventive Principle:

Principle #1Segmentation

3Ease of operation

If the system uses neural networks to automatically select speech recognition models, then the ease of operation improves, but the device complexity increases

Engineering Contradiction:

Improveautomatic model selectionVSAvoidneural network implementation

Core Design Contradiction:

Ease of operationVSDevice complexity

Solution Approach 1:

The system introduces a statistical classifier as an intermediary component that processes context information and audio characteristics to select the appropriate speech recognition model, providing automated operation without requiring complex neural network implementation

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS9311915B2Context-based speech recognition

Publication Date: 2016.04.12 GOOGLE LLC

AI summary

A processing system receives an audio signal encoding a portion of an utterance. The processing system receives context information associated with the utterance, wherein the context information is not derived from the audio signal or any other audio signal. The processing system provides, as input to a neural network, data corresponding to the audio signal and the context information, and generates a transcription for the utterance based on at least an output of the neural network.