ASR Controller Routing for Multi-Domain Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic speech recognition (ASR) systems are limited to specific subject matter domains, leading to recognition errors and communication breakdowns when attempting to connect with a wide range of domains using a single system.
Innovation Solution
The implementation of an ASR architecture that supports multiple speech recognition systems through a client-server framework, where an input-output ASR controller assigns an affinity status to recognition modules based on prior results or user requests, routing future utterances to specific modules for improved efficiency and consistency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single ASR system attempts to cover multiple subject matter domains, then the system's versatility is improved, but recognition accuracy deteriorates due to recognition errors
Solution Approach 1:
The patent segments the ASR system into multiple specialized recognition modules, each dedicated to a specific subject matter domain (e.g., hotel reservations, home automation). This segmentation allows each module to maintain high recognition accuracy for its domain while the overall system achieves broad domain coverage through the collection of specialized modules.
Solution Approach 2:
The controller serves as a universal component that manages multiple specialized recognition modules, routing incoming speech to the appropriate module based on domain identification. This multi-functional controller enables the system to handle diverse domains while maintaining the specialization benefits of individual modules.
2Reliability
If multiple speech recognition modules are maintained to cover different domains, then recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The controller acts as an intermediary between the user and multiple recognition modules. It manages the complexity by handling domain identification, module selection, and result aggregation, thereby shielding users from the underlying system complexity while enabling accurate multi-domain recognition.
Solution Approach 2:
The patent merges multiple specialized recognition modules under a single unified controller that coordinates their operation. This combining approach allows the system to leverage the accuracy benefits of specialized modules while presenting a unified interface and managing complexity centrally through the controller.
3Adaptability or versatility
If all recognition modules process every utterance to ensure domain coverage, then versatility is improved, but processing efficiency deteriorates due to increased processor loads and bandwidth usage
Solution Approach 1:
The controller performs preliminary domain identification and module selection before routing the speech utterance to recognition modules. This preliminary action ensures that only the necessary specialized module processes each utterance, avoiding redundant processing and maintaining high processing efficiency while preserving domain coverage capabilities.
Data Source
AI summary
A system and method to receive a spoken utterance and convert the spoken utterance into a recognized speech results through an automatic speech recognition service. A spoken utterance into a recognized speech result through an automatic speech recognition service. The recognized speech results are interpreted through a natural language processing module. A normalizer processes the recognized speech results that transforms the recognized speech interpretations into predefined form for a given automatic speech recognition domain and further determines which automatic speech recognition domains or the recognized speech results are processed by a dedicated dialogue management proxy module or a conversation module.