Speech Recognition Scoring via Keyword Parameter Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition methods are inefficient and prone to misinterpretation, often requiring users to rephrase their requests multiple times before being accurately understood, as they typically provide a single-round recognition that fails to account for variations in user intent and context.
Innovation Solution
A method and system for speech recognition that involves obtaining candidate recognition results, extracting keywords, determining parameters such as matching degrees, popularity, and user preferences, and generating updating coefficients to refine preliminary scores, ultimately selecting a target recognition result based on these analyses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If single-round recognition is used, then the system complexity is low, but the recognition accuracy is insufficient
Solution Approach 1:
The patent segments the speech recognition process into multiple independent rounds, where each round generates candidate results and scores. The system divides the recognition task into discrete stages (obtaining candidates, extracting keywords, determining parameters, updating scores) that can be executed sequentially, allowing complex multi-round processing while maintaining manageable system architecture through modular design.
Solution Approach 2:
The patent applies preliminary action by obtaining multiple candidate recognition results and their preliminary scores before final selection. The system performs preliminary keyword extraction, parameter determination, and score updating in advance of the final recognition result selection, allowing thorough analysis of multiple candidates rather than committing to a single immediate result.
2Measurement precision
If multiple candidate results are evaluated with multiple parameters, then the recognition accuracy improves, but the processing time increases
Solution Approach 1:
The patent implements periodic action through multi-round recognition where the system periodically processes candidate results through standardized stages. Each round follows a periodic pattern of obtaining candidates, extracting keywords, determining parameters, updating scores, and selecting results. This structured periodic processing allows efficient batch evaluation of multiple candidates rather than ad-hoc analysis, reducing overall processing time.
Solution Approach 2:
The patent utilizes parameter changes by dynamically adjusting recognition parameters (such as score thresholds, keyword weights, and parameter combinations) based on the evaluation progress and candidate characteristics. The system changes parameters during the multi-round process to optimize the balance between thorough evaluation and processing efficiency, adapting the evaluation criteria as more candidates are analyzed.
3Loss of information
If keyword extraction and parameter determination are performed for each candidate, then the understanding of user intent improves, but the computational complexity increases
Solution Approach 1:
The patent applies the extraction principle by isolating and extracting specific keywords from each candidate recognition result independently. The system extracts relevant keywords, determines their parameters, and evaluates them separately rather than analyzing the entire candidate text as a monolithic unit. This extraction approach reduces computational complexity by focusing processing on key informative elements rather than all text content.
Solution Approach 2:
The patent implements partial action by performing keyword extraction and parameter determination selectively on the most promising candidate results rather than exhaustively processing every single candidate with equal depth. The system may perform more detailed analysis on top candidates while using lighter processing for lower-ranked candidates, optimizing the balance between thorough intent understanding and computational resource utilization.
Data Source
AI summary
Systems and methods for speech recognition are provided. The method may include obtaining a plurality of candidate recognition results of speech information uttered by a user and a plurality of preliminary scores corresponding to the plurality of candidate recognition results, respectively. The method may further include, for each of the plurality of candidate recognition results, extracting one or more keywords from the candidate recognition result and determining at least one parameter associated with the one or more extracted keywords. The method may further include, for each of the plurality of candidate recognition results, generating an updating coefficient based on the at least one parameter and updating the preliminary score based on the updating coefficient to generate an updated score. The method may further include determining, from the plurality of candidate recognition results, a target recognition result based on the plurality of updated scores.