Dynamic Parse Rule Generation for Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Speech recognition systems face challenges in anticipating and parsing all possible user commands, leading to misinterpretation and inefficiencies in command processing due to the inability to generate all possible parsing rules beforehand.
Innovation Solution
A method that selects groups of queries and corresponding responses from log data, utilizing a high precision system to determine attribute types and generate parse rules, reducing the need for detailed studies and evaluator-generated command inputs, thereby reducing parse model maintenance costs and time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the speech recognition system attempts to generate all possible parsing rules beforehand to cover all user commands, then the command recognition accuracy would improve, but the system complexity and maintenance burden would increase significantly
Solution Approach 1:
The system transitions from static pre-defined parsing rules to dynamic rule generation. The parse model generator creates parsing rules on-demand based on actual user queries and responses, allowing the system to adapt to new command patterns without manual intervention and reducing maintenance burden while maintaining accuracy
Solution Approach 2:
The system enables self-service by automatically generating parsing rules from user interactions. The parse model generator uses actual queries and responses to create new parsing rules autonomously, eliminating the need for continuous manual rule creation and updating, thus reducing complexity while improving coverage
2Measurement precision
If the system manually creates parsing rules through detailed studies and evaluator-generated commands, then the parse accuracy improves, but the time and cost for parse model maintenance increases
Solution Approach 1:
The system performs preliminary action by collecting and analyzing actual user queries and responses in advance. The parse model generator uses this accumulated data to proactively generate parsing rules before they are needed, ensuring accuracy while eliminating the time-consuming manual rule creation process
Solution Approach 2:
The system implements feedback by using actual user queries and responses to generate new parsing rules. The parse model generator continuously learns from real user interactions, improving parse accuracy over time while automatically reducing maintenance time through data-driven rule generation
3Adaptability or versatility
If the speech recognition system uses a high recall service to capture all possible commands, then the command coverage improves, but the precision of command interpretation decreases
Solution Approach 1:
The system segments the command processing into two stages: first, a high-recall service captures all possible commands broadly; second, the parse model generator refines these using actual query-response data to improve interpretation precision. This segmentation allows both broad coverage and accurate interpretation to coexist
Data Source
AI summary
A language processing system collects similar queries and respective responses and aggregated by responses. Incorrect responses are determined and filtered by the aggregation. The remaining responses are then used to query a high precision system for attributes of entities specified by the queries. The attribute type is determined from the responses of the high precision system, and corresponding parse rules are generated. The parse rules are then associated with an operation that yields a response that specifies an attribute of the attribute type.