Context-Aware ASR Word Boosting for Command Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated speech recognition (ASR) systems face accuracy issues when boosting specific words or phrases, as they often over-boost these words, leading to incorrect transcriptions and degrading overall recognition accuracy for other utterances.
Innovation Solution
The system adjusts language model scores for target words by considering the scores of surrounding n-grams, applying a boosting factor that enhances the recognition of specific words while maintaining accuracy for other phrases, and dynamically adjusts scores based on context or permanently for enhanced ASR performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If word boosting is applied to enhance recognition of specific commands, then recognition accuracy for those commands improves, but overall recognition accuracy for other utterances degrades
Solution Approach 1:
The patent applies different boosting strategies to different words based on their individual characteristics. Instead of uniformly boosting all target words, the system analyzes each word's context, frequency, and importance to determine appropriate boosting levels, thereby improving specific command recognition while minimizing impact on other utterances
Solution Approach 2:
The system dynamically adjusts language model parameters (such as word probabilities and n-gram scores) based on the specific boosting requirements. By modifying these parameters selectively for different words and contexts, the system enhances command recognition accuracy while maintaining overall speech recognition reliability
2Measurement precision
If boosting factor is increased to improve specific word recognition, then recognition of target words improves, but transcription accuracy for non-target words degrades
Solution Approach 1:
The patent applies partial boosting rather than excessive uniform boosting across all words. By selectively applying boosting factors only to relevant target words and their immediate contexts, the system achieves improved target word recognition while avoiding the degradation of transcription accuracy for non-target words that would result from excessive global boosting
Data Source
AI summary
Systems and methods are provided for performing automated speech recognition. The systems and methods perform operations comprising: accessing a language model that includes a plurality of n-grams, each of the plurality of n-grams comprising a respective sequence of words and corresponding LM score; selecting a target word to boost in the language model; receiving a boosting factor for the target word; identifying a target n-gram in the language model that includes the target word; identifying a subset of n-grams of the plurality of n-grams that include words in a portion of the target n-gram; and adjusting the LM score of the target n-gram based on the LM scores of the subset of n-grams and the boosting factor.