Speech Recognition Wildcard Threshold Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems in inventory management face inefficiencies due to high acceptance thresholds that lead to unnecessary rejections of correct speech with low confidence scores, reducing productivity and accuracy.
Innovation Solution
A speech recognition system that adjusts acceptance thresholds using expected responses with wildcard words, allowing for more generalized matching and adaptation of models based on hypothesized words that match either expected words or wildcard words, thereby improving recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a high acceptance threshold is used in speech recognition, then false rejections of correct speech are reduced, but correct speech with low confidence scores is unnecessarily rejected, reducing productivity
Solution Approach 1:
The patent applies dynamics by making the acceptance threshold adjustable rather than fixed. The threshold is dynamically modified based on whether the hypothesized speech matches an expected response pattern. When a match is found, the threshold is lowered to accept speech with lower confidence scores, thereby reducing unnecessary rejections and improving productivity while maintaining reliability when matches are not found.
2Productivity
If the acceptance threshold is lowered to accept more speech inputs, then productivity improves, but false rejections increase and recognition accuracy decreases
Solution Approach 1:
The patent applies local quality by applying different acceptance thresholds to different speech inputs based on their characteristics. Instead of using a uniform threshold for all speech, the system identifies speech that matches expected response patterns and applies a lower threshold specifically to those cases, while maintaining a higher threshold for other speech. This localized adjustment improves productivity for expected responses without compromising overall recognition accuracy.
3Reliability
If expected responses are limited to known complete responses, then speech recognition accuracy is maintained, but the system cannot handle partial knowledge or variable responses, reducing adaptability
Solution Approach 1:
The patent applies universality by enabling the expected response mechanism to handle multiple types of responses through wildcard patterns. The system can match complete expected responses as well as partial responses containing wildcards that represent variable portions. This allows a single expected response definition to cover multiple possible valid responses, increasing adaptability while maintaining recognition accuracy through the matching mechanism.
Data Source
AI summary
A speech recognition system used in a workflow receives and analyzes speech input to recognize and accept a user's response to a task. Under certain conditions, a user's response might be expected. In these situations, the expected response may modify the behavior of the speech recognition system to improve recognition accuracy. For example, if the hypothesis of a user's response matches the expected response then there is a high probability that the user's response was recognized correctly. An expected response may include expected words and wildcard words. Wildcard words represent any recognized word in a user's response. By including wildcard words in the expected response, the speech recognition system may make modifications based on a wide range of user responses.