Voice Search Confidence Measure Generator for Speech Recognition Errors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice search systems face challenges in generating confidence measures due to high automatic speech recognition error rates and linguistic diversity, which are exacerbated by large vocabularies and vast search spaces, making it difficult for existing technologies to accurately interpret user queries and provide robust results.
Innovation Solution
A method is introduced for generating a confidence measure in voice search systems by selecting relevant features from voice search components and training a model using a computer processor to produce a voice search confidence measure, incorporating a smoothed n-gram language model and vector space model to enhance robustness to errors and diversity, and using a maximum entropy classifier to assign probabilities to query results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If voice search systems use large vocabularies and vast search spaces to handle linguistic diversity, then the system's adaptability improves, but speech recognition error rates increase
Solution Approach 1:
The patent introduces a confidence measure generator as an intermediary component between the speech recognizer and the dialog manager. This mediator evaluates multiple features (ASR confidence scores, search result quality metrics, linguistic feature matches) to compensate for high error rates, enabling the system to maintain reliability despite large vocabulary and linguistic diversity challenges
2Adaptability or versatility
If the vocabulary size is increased to cover more user queries, then the system's versatility improves, but the complexity of generating accurate confidence measures worsens
Solution Approach 1:
The confidence measure generation process is segmented into multiple independent feature evaluation components: ASR confidence scoring, search result quality assessment, and linguistic feature matching. Each component processes specific aspects separately, then their results are combined to form the overall confidence measure, reducing the complexity of handling large vocabularies
3Ease of manufacture
If existing confidence measure methods from speech recognition are applied directly to voice search, then implementation simplicity improves, but measurement precision worsens due to insufficient training data
Solution Approach 1:
The patent creates a universal confidence measure framework that integrates multiple data sources and evaluation criteria beyond traditional speech recognition methods. It combines ASR confidence scores with search-specific metrics (result quality, query matching) and linguistic features, making the system adaptable to voice search's unique challenges while maintaining implementation feasibility
Data Source
AI summary
A method of generating a confidence measure generator is provided for use in a voice search system, the voice search system including voice search components comprising a speech recognition system, a dialog manager and a search system. The method includes selecting voice search features, from a plurality of the voice search components, to be considered by the confidence measure generator in generating a voice search confidence measure. The method includes training a model, using a computer processor, to generate the voice search confidence measure based on selected voice search features.