Speech Recognition Error Correction via Pinyin Intermediary
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Speech recognition results often contain errors due to limited accuracy, which hinders effective understanding and requires correction to improve reliability.
Innovation Solution
A method involving phonetic notation to convert speech recognition results into pinyin, followed by obtaining candidate texts through a decoding network based on a pronunciation dictionary and language models, with the optimum candidate text being determined and corrected if it meets preset conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If speech recognition is performed to convert speech to text, then the speech understanding process can proceed, but errors in the recognition results reduce the accuracy of speech understanding
Solution Approach 1:
The patent introduces pinyin as an intermediary representation between the speech recognition result and the candidate texts. The correction system converts the recognition result to pinyin, uses the pinyin to query a pronunciation dictionary for candidate texts, and then selects the most appropriate candidate. This intermediary approach allows the system to leverage phonetic information to improve accuracy without requiring direct speech-text mapping
Solution Approach 2:
The patent implements a feedback mechanism where the speech recognition result is corrected by comparing it against candidate texts derived from pinyin conversion. The system evaluates multiple candidates and selects the optimum one, providing feedback to improve the overall speech understanding accuracy. This feedback loop addresses errors by using phonetic matching to identify and correct mismatches
2Measurement precision
If phonetic notation and candidate text generation are performed to correct errors, then the accuracy of speech understanding is improved, but the system complexity increases
Solution Approach 1:
The patent performs phonetic notation conversion to pinyin as a preliminary action before candidate text generation and selection. By converting the recognition result to pinyin first, the system prepares the data in a form that can be efficiently queried against the pronunciation dictionary. This preliminary processing simplifies the subsequent candidate generation step by establishing a clear phonetic-matching pathway
Solution Approach 2:
The patent creates a phonetic copy of the speech recognition result by converting it to pinyin. This pinyin copy serves as a bridge to query the pronunciation dictionary and generate candidate texts. Instead of directly manipulating the original recognition result, the system works with this phonetic representation, which simplifies the correction process by decoupling the error correction logic from the original text structure
Data Source
AI summary
A method, a device, and a storage medium for correcting an error in a speech recognition result are provided. The method includes: performing phonetic notation on a speech recognition result to be corrected, to obtain a pinyin corresponding to the speech recognition result; obtaining one or more candidate texts according to the pinyin, and determining an optimum candidate text from the one or more candidate texts; judging whether the optimum candidate text satisfies a preset condition; and determining the optimum candidate text as a corrected result of the speech recognition result to be corrected in response to satisfying the preset condition.