Spoken Dialog Agent System for Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice-controlled home electric appliances face limitations due to restricted local storage capacity, leading to delayed responses as they rely on cloud servers for vocabulary and expression handling, resulting in noticeable delays in spoken dialog interactions.
Innovation Solution
A spoken dialog agent system that recommends local-side speech phrases for immediate device control, utilizing a cloud server to match user inputs with synonyms stored in a cloud dictionary, thereby reducing latency and improving response times by shifting processing to locally registered phrases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If speech recognition is performed using a cloud server with a large vocabulary dictionary, then the vocabulary coverage and expression handling capability are improved, but the response time increases due to communication latency
Solution Approach 1:
The patent divides the dictionary into two segments: a local dictionary stored in the terminal device with limited vocabulary, and a cloud dictionary stored in the cloud server with comprehensive vocabulary. The system first attempts matching against the local dictionary for rapid response, and only when that fails does it query the cloud dictionary, thus segmenting the recognition process to balance speed and coverage
Solution Approach 2:
The system performs preliminary action by pre-storing a subset of commonly used vocabulary and expressions in the local dictionary of the terminal device. This allows the device to handle routine speech commands locally without cloud communication, preparing the system in advance for fast local matching while maintaining the option to access the full cloud dictionary when needed
2Measurement precision
If a local-side device stores a comprehensive vocabulary dictionary, then the speech recognition accuracy is improved, but the storage capacity requirement exceeds the device's limited local storage
Solution Approach 1:
The patent segments the comprehensive vocabulary dictionary into two parts: a condensed local dictionary that fits within the terminal device's limited storage capacity, and a complete cloud dictionary that resides on the cloud server. This segmentation allows the device to maintain high recognition accuracy for common phrases locally while relying on the cloud for comprehensive vocabulary coverage without exceeding local storage constraints
Solution Approach 2:
The cloud server acts as a universal resource that serves multiple terminal devices, providing comprehensive vocabulary storage and updating capabilities that individual devices cannot achieve alone. This multi-functionality allows each device to have access to extensive vocabulary through the cloud without each device needing to store the complete dictionary locally
3Adaptability or versatility
If speech commands are processed through cloud server communication, then the vocabulary diversity is improved, but the dialog interaction speed decreases due to round trip time
Solution Approach 1:
The system performs preliminary action by pre-loading essential vocabulary and speech patterns into the local dictionary during device initialization or updates. This allows the device to immediately process common speech commands using locally stored vocabulary without waiting for cloud communication, thus maintaining high dialog speed for routine interactions while preserving access to diverse vocabulary through the cloud when needed
Solution Approach 2:
The patent segments the speech processing workflow into two paths: a fast local processing path for vocabulary present in the local dictionary, and a slower cloud processing path for vocabulary only available in the cloud dictionary. This segmentation enables the system to maintain high overall dialog speed by handling the majority of common interactions locally while still providing access to diverse vocabulary through selective cloud queries
Data Source
Figure 1A~1C
Figure 2
Figure 3
AI summary
An information processing method includes: acquiring first voice information indicating a voice of a user input from a microphone; outputting, to a server via a network, first text string information generated from the first voice information, when the first text string information does not match any of pieces of text string information in the first database; acquiring, from the server, first semantic information and/or a control command corresponding to the first semantic information, when a second database includes a piece of text string information matched with the first text string information and the matched piece of text string information is associated with the first semantic information therein; instructing at least one device to execute an operation based on the first semantic information and/or the control command; and outputting, to a speaker, second voice information generated from second text string information, the second text string information being registered and associated with the first semantic information in the first database.