Speech Recognition Confidence via Dialog Context Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Speech recognition systems face challenges in determining recognition confidence due to variations in dialects and speech clarity, as they typically do not consider the context of overall speech dialog events when evaluating recognition accuracy.
Innovation Solution
A method that identifies features from both the current speech utterance and previous dialog events to calculate a confidence score, using a combination of local and global features to improve recognition confidence by leveraging context from the entire speech dialog.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only local features from the current speech utterance are used to determine recognition confidence, then the calculation process is simple and fast, but the recognition accuracy is limited due to lack of contextual information
Solution Approach 1:
The patent transitions from analyzing only local features (single utterance) to incorporating global features (entire speech dialog) as an additional dimension of analysis. This multi-dimensional approach combines local utterance characteristics with global dialog context, enabling more accurate recognition confidence determination without requiring complete re-analysis of all speech data.
Solution Approach 2:
The system performs preliminary analysis of global dialog features before final confidence determination. By pre-identifying and storing relevant global features from the speech dialog, the system prepares contextual information in advance that can be quickly integrated with local utterance features during confidence calculation, reducing real-time computational burden.
2Measurement precision
If global features from the entire speech dialog are incorporated into confidence calculation, then recognition accuracy improves through contextual understanding, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the speech dialog into distinct local utterances and global dialog components, allowing independent analysis of each segment. Local features are extracted from individual utterances while global features are derived from the overall dialog structure. This segmentation enables parallel processing and selective integration of features, reducing the time required for comprehensive confidence calculation.
Solution Approach 2:
The system extracts only the most relevant global features from the speech dialog that have significant impact on recognition confidence, rather than analyzing all dialog data in full detail. By selecting and extracting key contextual indicators, the system maintains high accuracy while minimizing computational overhead and processing time.
Data Source
AI summary
On a computing device a speech utterance is received from a user. The speech utterance is a section of a speech dialog that includes a plurality of speech utterances. One or more features from the speech utterance are identified. Each identified feature from the speech utterance is a specific characteristic of the speech utterance. One or more features from the speech dialog are identified. Each identified feature from the speech dialog is associated with one or more events in the speech dialog. The one or more events occur prior to the speech utterance. One or more identified features from the speech utterance and one or more identified features from the speech dialog are used to calculate a confidence score for the speech utterance.