Sample acquisition method and apparatus, electronic device, storage medium, and program product
By analyzing the phonetic differences of polyphonic characters, samples that are difficult to predict were screened out, and the training of the content recognition model was optimized. This solved the problem of insufficient accuracy in polyphonic character recognition and improved the model's recognition accuracy and reliability.
CN122347951APending Publication Date: 2026-07-07TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by
Patent Information
- Application Number
- CN202510021257.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-06
- Publication Date
- 2026-07-07
AI Technical Summary
Technical Problem
Existing neural network models struggle to fully cover all possible contextual variations when dealing with polyphonic characters, resulting in insufficient accuracy and reliability in polyphonic character recognition.
Method used
By acquiring original text containing polyphonic characters, performing speech-to-text conversion, analyzing phonetic differences, filtering out samples with difficult polyphonic character prediction, and optimizing the training process of the content recognition model.
Benefits of technology
This improved the accuracy and reliability of the content recognition model in recognizing polyphonic characters, reduced misrecognition during speech transcription, and enhanced the model's recognition performance.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122347951A_ABST
Abstract
The application discloses a sample acquisition method and device, electronic equipment, storage medium and program product; the embodiment of the application can acquire at least one original text containing a polyphonic word, and determine a target position of the polyphonic word in the original text; the original text is subjected to speech conversion processing to obtain speech data; the speech data is subjected to text conversion processing to generate a transcription text; according to the target position of the polyphonic word in the original text, the original text and the transcription text are subjected to phonetic alphabet difference analysis processing to obtain phonetic alphabet difference information of the polyphonic word in the original text and the transcription text; based on the phonetic alphabet difference information, a target text is selected from the original text, and the target text is used for identifying and training the polyphonic word of a content recognition model. In the embodiment of the application, the content recognition model is trained by screening out the polyphonic word sample with recognition errors. Therefore, the scheme can improve the recognition accuracy and reliability of the content recognition model for the polyphonic word.
Need to check novelty before this filing date? Find Prior Art