Speech Recognition Pairwise Hypothesis Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional virtual assistant applications face challenges in accurately generating textual representations of user spoken utterances due to factors like pitch variations in noisy environments and differences in tone between speakers, which are not adequately addressed by existing systems.
Innovation Solution
A computer-implemented method and system that generates textual representations of user spoken utterances by receiving the utterance, generating multiple hypotheses, pairing them, determining pair scores using a pairwise classifier, and ranking them based on both pair scores and user-specific and acoustic features to select the most accurate representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional virtual assistant applications use traditional speech recognition methods, then the system complexity remains low, but the accuracy of textual representation generation deteriorates in noisy environments and with pitch variations
Solution Approach 1:
The speech recognition system is divided into multiple independent components: acoustic model, language model, and pairwise classifier. Each component processes specific aspects of speech recognition independently, allowing for improved accuracy through specialized processing without requiring complete system redesign.
Solution Approach 2:
The patent introduces a new dimension of analysis by comparing hypotheses in pairs rather than evaluating them individually. This pairwise comparison approach adds a relational dimension to the recognition process, enabling the system to resolve ambiguities that single-hypothesis methods cannot address.
2Measurement precision
If the system generates multiple hypotheses for spoken utterances, then the accuracy of textual representation improves, but the processing time increases
Solution Approach 1:
The system generates multiple hypotheses beyond what a single correct answer would require, then uses pairwise comparison to efficiently eliminate incorrect options. This partial action approach processes only the necessary comparisons to achieve high accuracy without exhaustively evaluating all possible hypotheses.
Solution Approach 2:
The patent changes the evaluation parameter from individual hypothesis scoring to pairwise hypothesis comparison. This parameter transformation allows the system to leverage relative differences between hypotheses, improving accuracy while maintaining computational efficiency through focused comparison rather than comprehensive evaluation.
3Measurement precision
If the system adapts to individual speaker tones and characteristics, then the measurement precision improves, but the device complexity increases due to additional processing requirements
Solution Approach 1:
The system performs preliminary processing by generating multiple candidate hypotheses that account for speaker characteristics and tone variations before the pairwise comparison stage. This preliminary action prepares the data in advance, allowing the main recognition algorithm to focus on discrimination rather than adaptation, thereby improving accuracy without proportionally increasing complexity.
Data Source
AI summary
A method and a system for generating textual representation of user spoken utterance is disclosed. The method comprises receiving an indication of the user spoken utterance; generating, at least two hypotheses; generating, by the electronic device, from the at least two hypotheses a set of paired hypotheses, a given one of the set of paired hypotheses including a first hypothesis paired with a second hypothesis; determining, for the given one of the set of paired hypotheses, a pair score; generating a set of utterance features, the set of utterance features being indicative of one or more characteristics associated with the user spoken utterance; ranking, the first hypothesis and the second hypothesis based at least on the pair score and the set of utterance features; and in response to the first hypothesis being a highest ranked hypothesis, selecting the first hypothesis as the textual representation of user spoken utterance.