False Accept Detection Model for Speech Intent Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech processing systems in devices like Amazon Alexa and Google Home often incorrectly determine user intent due to limitations in computing power and knowledge, leading to incorrect routing of dialog to network-based domains, resulting in a frustrating user experience and decreased performance of false accept detection models.
Innovation Solution
A false accept detection model using a long short-term memory (LSTM) artificial recurrent neural network architecture evaluates user utterances across domains to confirm intent, combining word-level and character-level features from automatic speech recognition and natural language understanding outputs, and applies a secondary analysis to adjust confidence scores and prevent incorrect routing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the speech processing component uses basic automatic speech recognition and natural language understanding modules to determine user intent, then the device can perform simple routing tasks, but the intent determination accuracy is insufficient leading to incorrect routing decisions
Solution Approach 1:
The patent introduces a false accept detection model as an intermediary component between the speech processing component and network-based domains. This model receives utterances from the speech processing component and performs secondary analysis to verify intent determination accuracy. By adding this intermediate verification layer, the system achieves higher routing accuracy without requiring the speech processing component itself to become significantly more complex.
Solution Approach 2:
The patent segments the intent determination process into two independent stages: primary intent determination by the speech processing component using ASR and NLU modules, and secondary verification by the false accept detection model. This segmentation allows each component to specialize in its specific function, maintaining simplicity at the speech processing level while achieving high overall accuracy through the specialized verification model.
2Speed
If the speech processing component makes irreversible routing decisions based on limited computing power and knowledge, then routing speed is maintained, but incorrect routing occurs when intent determination is wrong
Solution Approach 1:
The false accept detection model performs preliminary verification of intent determination before the final routing decision is executed. By analyzing utterances in advance and identifying potential false accepts, the system can prevent incorrect routing decisions without significantly delaying the overall process. The model is trained to quickly identify suspicious cases that require further review.
Solution Approach 2:
The system implements a feedback mechanism where the false accept detection model analyzes routing decisions and provides corrective feedback when errors are detected. The model compares its analysis with the speech processing component's intent determination and can override incorrect routing decisions. This feedback loop continuously improves routing reliability while maintaining operational speed.
3Measurement precision
If the false accept detection model is updated to improve accuracy, then model performance increases, but regression occurs when speech models or traffic patterns change
Solution Approach 1:
The false accept detection model is designed with dynamic adaptability to adjust to changing speech models and traffic patterns. The model can be retrained and updated without requiring complete system redesign, allowing it to adapt to new speech recognition capabilities and evolving user behavior patterns. This dynamic nature maintains detection accuracy while accommodating system changes.
Solution Approach 2:
The system employs parameter adjustment mechanisms that allow the false accept detection model to adapt to changes in speech model characteristics and traffic patterns. By modifying model parameters and retraining with updated data, the system maintains high detection accuracy even when underlying speech recognition models or user interaction patterns change over time.
Data Source
AI summary
A new model is introduced into a particular domain that receives a routing of a dialog from a speech processing component. A method associated with the model includes running a set of test utterances through the speech processing component that enables a spoken language dialog with a user to establish a base line score associated with processing for the set of test utterances. The speech processing component determines an intent of the user and routes the spoken language dialog to a network-based domain based on the intent. The method includes establishing an automatic test run of the set of test utterances to obtain a current score and, when a threshold associated with a difference between the current score and the base line score is breached, switching, at the network-based domain, from the false accept detection model to a second model.