Speech Recognition Mapping Function for Homophone Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition methods struggle with high recognition costs and low accuracy when dealing with homophones, polyphones, or ambiguous pronunciation, as they typically recognize either Chinese character or Pinyin results, leading to poor user experience and increased misunderstanding.
Innovation Solution
A method and device that set initial Chinese character and Pinyin coefficients to generate mapping functions, train them using preset samples, and calculate a joint loss function to produce a target mapping function, allowing for simultaneous Chinese character and Pinyin recognition without separate requests, reducing costs and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If two separate recognition methods (Chinese character recognition and Pinyin recognition) are used to recognize speech containing homophones, polyphones or ambiguous pronunciation, then the recognition accuracy is improved, but the recognition cost increases
Solution Approach 1:
The patent combines Chinese character recognition and Pinyin recognition into a single unified recognition system. The speech input is processed once through integrated recognition modules that simultaneously generate both Chinese character results and Pinyin results, eliminating the need for separate recognition requests and reducing computational overhead while maintaining high recognition accuracy for homophones and polyphones
Solution Approach 2:
The recognition system is designed to perform multiple functions simultaneously - it can recognize speech and output both Chinese character results and Pinyin results in a single processing pass. This multi-functional capability allows the system to handle different recognition needs (character-based or Pinyin-based) without requiring separate specialized systems, thereby reducing overall recognition cost while preserving accuracy
2Device complexity
If only one recognition method (either Chinese character or Pinyin) is used, then the recognition cost is reduced, but the recognition accuracy decreases for speech containing homophones, polyphones or ambiguous pronunciation
Solution Approach 1:
The system merges Chinese character recognition and Pinyin recognition capabilities into a single integrated framework. By processing speech input once through both recognition pathways simultaneously, the system achieves the accuracy benefits of dual-method recognition while incurring only the cost of a single recognition operation, thus resolving the contradiction between cost and accuracy
3Loss of information
If two separate recognition requests are sent to obtain both Chinese character and Pinyin results, then the completeness of recognition results is improved, but the processing time increases
Solution Approach 1:
The system implements continuous parallel processing where Chinese character recognition and Pinyin recognition occur simultaneously in a single processing pipeline. Both recognition processes share common preprocessing stages and operate concurrently, ensuring that both types of results are generated in one continuous operation rather than through sequential requests, thereby reducing total processing time while maintaining result completeness
Data Source
AI summary
A method for recognizing speech comprises: respectively setting initial values of a Chinese character coefficient and a Pinyin coefficient, generating a Chinese character mapping function according to the initial value of the Chinese character coefficient, and generating a Pinyin mapping function according to the initial value of the Pinyin coefficient (S101); training the Chinese character mapping function and the Pinyin mapping function using a plurality of preset training samples, calculating training results as parameters of a joint loss function, and generating a target mapping function according to calculation results (S102); and recognizing, according to the target mapping function, speech to be recognized, so as to obtain a Chinese character recognition result and a Pinyin recognition result of the speech to be recognized (S103). The method reduces the cost of speech recognition while ensuring the accuracy of speech recognition.