Integrated Shared Dictionary for Dialog Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dialog systems face challenges in reducing memory and processing resources, making it difficult to implement them on low-power devices due to the separate resource requirements of automated speech recognition (ASR) and natural language understanding (NLU) components.
Innovation Solution
Implementing an integrated shared dictionary that both ASR and NLU subsystems can use, which reduces memory footprint and processing time by allowing for shared resources and streamlined processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If separate resources are allocated for ASR and NLU components, then each component can perform its tasks independently, but memory usage and processing resources increase significantly
Solution Approach 1:
The patent merges the separate dictionary resources of ASR and NLU components into a single shared dictionary. The ASR component's pronunciation dictionary and the NLU component's word meaning dictionary are combined into one unified data structure that both components can access simultaneously, eliminating redundant storage of word-pronunciation-meaning triplets.
Solution Approach 2:
The shared dictionary serves multiple functions: it acts as the pronunciation reference for ASR speech-to-text conversion, simultaneously serves as the word meaning database for NLU text-to-meaning conversion, and provides a common vocabulary reference for both components. This multi-functional resource reduces overall system memory requirements while maintaining independent operational capabilities of ASR and NLU.
2Reliability
If separate dictionaries are maintained for ASR and NLU, then each component has dedicated resources, but processing time increases due to redundant lookups
Solution Approach 1:
The patent combines separate dictionary lookup operations into a single unified lookup process. Instead of ASR performing a lookup in its pronunciation dictionary and then NLU performing another lookup in its word meaning dictionary for the same word, both components access the shared dictionary simultaneously, performing only one physical lookup operation per word.
Solution Approach 2:
The shared dictionary is pre-organized with a unified structure that allows both ASR and NLU to access word information efficiently in a single operation. The dictionary entries are structured to provide pronunciation, part-of-speech, and meaning information together, so that preliminary organization enables rapid simultaneous access by both components without sequential processing delays.
3Adaptability or versatility
If independent components are used for ASR and NLU, then system modularity is maintained, but device complexity increases making implementation on low-power devices difficult
Solution Approach 1:
The patent merges the data storage layer (dictionaries) while maintaining the operational independence of ASR and NLU components. The architecture combines the separate dictionary files into one shared data structure, but ASR and NLU remain separate processing modules that independently access this shared resource, thus reducing complexity without sacrificing modularity.
Solution Approach 2:
The system is segmented into distinct functional components (ASR, NLU, and the shared dictionary as a common resource layer). This segmentation allows each component to be developed, optimized, and maintained independently while sharing common vocabulary data, making the system adaptable to different hardware platforms including low-power devices.
Data Source
AI summary
Techniques for reducing memory and processing resources used by a dialog system by sharing resources between pipelined processes of the dialog system. An integrated shared dictionary is constructed for concurrent use by automated speech recognition (ASR) and natural language understanding (NLU) subsystems of the dialog system. The integrated shared dictionary comprises multiple entries, with each entry comprising first information that is used by the ASR subsystem, second information used by the NLU subsystem, and information correlating the first information and the second information. The ASR subsystem uses the integrated shared dictionary to identify a dictionary entry containing a set of words corresponding to speech input. The dictionary entry information is communicated to the NLU subsystem, which uses the entry to generate a meaning representation for the speech input. Sharing the dictionary between the subsystems results in significant savings in memory resources used by the dialog system and speeds processing.