Speech Recognition Dictionary Compilation Using Virtual Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition dictionary compilation methods are costly and labor-intensive, struggle with recognizing speech with unique styles, phonetically similar words, composite words, and fail to efficiently correct recognition errors, often requiring high-quality text data and manual intervention.
Innovation Solution
A speech recognition dictionary compilation system that uses a virtual speech recognition processing section to generate predictive text data from analyzed text data, comparing it with actual speech recognition results to update dictionaries and language models, incorporating an acoustic model to account for phonetic resemblance and reduce recognition errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual correction and high-quality text data collection are used, then speech recognition accuracy is improved, but cost and labor intensity increase
Solution Approach 1:
The system automatically generates virtual speech recognition results by processing text data through the same speech recognition pipeline, eliminating the need for manual correction. The virtual results serve themselves as the ground truth for dictionary updates, replacing manual intervention with automated self-service processing.
Solution Approach 2:
The system creates virtual copies of speech recognition results by simulating the recognition process on text data. These virtual results replicate the structure and characteristics of actual speech recognition outputs, allowing the system to learn from synthetic data rather than requiring expensive manual annotation.
2Extent of automation
If statistical techniques are used to detect unknown words, then dictionary compilation is automated, but recognition errors for phonetically similar words persist
Solution Approach 1:
The system implements feedback by comparing virtual speech recognition results with actual text data to identify recognition errors. This feedback loop automatically detects when phonetically similar words are confused and uses the information to update the dictionary and language models, ensuring that automation does not perpetuate recognition errors.
3Manufacturing precision
If text data is processed to extract recognition errors, then language models are updated, but processing time and computational cost increase
Solution Approach 1:
The system performs preliminary action by generating virtual speech recognition results in advance before actual recognition occurs. By pre-processing text data through the speech recognition pipeline to create virtual results, the system prepares correction data ahead of time, reducing the computational burden during actual speech recognition tasks.
Data Source
AI summary
A speech recognition dictionary compilation assisting system can create and update speech recognition dictionary and language models efficiently so as to reduce speech recognition errors by utilizing text data available at a low cost. The system includes speech recognition dictionary storage section 105, language model storage section 106 and acoustic model storage section 107. A virtual speech recognition processing section 102 processes analyzed text data generated by the text analyzing section 101 by making reference to the recognition dictionary, language models and acoustic models so as to generate virtual text data resulted from speech recognition, and compares the virtual text data resulted from speech recognition with the analyzed text data. The update processing section 103 updates the recognition dictionary and language models so as to reduce different point(s) between both sets of text data.