Voice Recognition Error Correction via Pinyin Candidate Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Chinese speech recognition systems face significant accuracy reductions due to variations in personal pronunciations and homophones, leading to errors in place name recognition, particularly in vehicle navigation scenarios where incorrect place names are often generated.
Innovation Solution
A method and system for error correction in speech recognition that involves pinyin conversion, filtering, and Cartesian product operations to generate candidate place names, utilizing a similarity dictionary and pinyin dictionary to split and combine pinyin lists, and constructing queries based on weighted candidate sets to query an inverted index for accurate results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If speech recognition is used to convert speech into text, then the conversion process is automated, but recognition accuracy deteriorates due to personal pronunciation differences and homophones
Solution Approach 1:
The patent segments the speech recognition process into multiple stages: initial speech-to-text conversion, pinyin generation, candidate place name generation through Cartesian product operations, filtering, and final correction. This segmentation allows each stage to address specific errors, improving overall accuracy while maintaining automation.
Solution Approach 2:
The patent implements feedback mechanisms by generating candidate place names based on recognized speech, filtering these candidates through multiple criteria, and selecting the most probable match. This feedback loop corrects recognition errors by comparing against known place name databases and user preferences.
2Measurement precision
If multiple candidate place names are generated to improve accuracy, then recognition precision improves, but system complexity increases due to multiple processing steps
Solution Approach 1:
The patent performs preliminary actions by pre-generating pinyin candidates and place name candidates before final selection. The Cartesian product operations and filtering steps are executed in advance to create a ranked list of candidates, making the final selection process simpler and more accurate.
3Adaptability or versatility
If speech recognition is applied in vehicle navigation with large number of place names, then service coverage improves, but error rate increases due to homophones and pronunciation variations
Solution Approach 1:
The patent changes parameters by considering multiple pinyin variations and pronunciation patterns when generating candidate place names. It adjusts the candidate generation process to account for different pronunciation styles, thereby maintaining high accuracy across diverse place names and pronunciation variations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are a method and system for error correction in speech recognition. The method comprises: extracting a place name by means of speech recognition and semantic understanding, and converting the place name into place name pinyin (S100); splitting the place name pinyin to obtain N pinyin lists, and performing a first Cartesian product operation on the N pinyin lists to generate a first pinyin candidate set, where N is a natural number (S200); filtering the pinyin in the first pinyin candidate set to eliminate irregular pinyin (S300); performing a second Cartesian product operation on the pinyin output in the filtering step to obtain a second pinyin candidate set (S400); constructing a query based on the second pinyin candidate set (S500); and querying the query in an inverted index to obtain candidate place names with high to low probabilities and display the candidate place names to a user (S600).