Hybrid Intent Detection System for Voice Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing conversation processing devices rely on either handwritten domain ontologies and semantic grammars or sequence models, which may not effectively balance precision and recall in intent identification and slot filling, limiting their adaptability to different domains.
Innovation Solution
A hybrid rule-based and machine learning system that performs intent identification and slot filling using ranked semantic relationships, combining expert knowledge with machine-learned rules to tailor performance metrics such as precision and recall based on domain requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If handwritten domain ontologies and semantic grammars are used for intent identification, then precision may be improved, but recall deteriorates and adaptability to different domains is limited
Solution Approach 1:
The patent merges rule-based intent identification (using hand-written domain ontologies and semantic grammars) with machine learning-based intent identification (using sequence models trained on dialogue data). This hybrid approach combines the precision of rule-based methods with the adaptability and recall of machine learning methods, allowing the system to achieve high precision in specific domains while maintaining good recall and adaptability across different domains through the machine learning component.
2Adaptability or versatility
If sequence models are used for intent identification, then recall and adaptability are improved, but precision deteriorates
Solution Approach 1:
The system combines sequence models with rule-based methods where the sequence model provides initial intent candidates with good recall and adaptability, while the rule-based component (domain ontologies and semantic grammars) refines these candidates to improve precision. The hybrid architecture allows the strengths of both approaches to complement each other.
Solution Approach 2:
The patent introduces an intermediary component that bridges the sequence model and final intent determination. This intermediary uses domain ontologies and semantic grammars to filter and refine the outputs from the sequence model, acting as a mediator that improves precision without completely sacrificing the adaptability and recall benefits of the machine learning approach.
3Measurement precision
If rule-based methods are used for slot filling, then precision is improved, but recall and productivity deteriorate
Solution Approach 1:
The patent merges rule-based slot filling with machine learning-based slot filling. The rule-based component ensures precision by applying domain-specific constraints and validation, while the machine learning component (sequence models) handles the bulk of slot extraction efficiently, improving overall productivity. This hybrid approach maintains high precision while significantly reducing processing time compared to purely rule-based systems.
Data Source
AI summary
A system determines user intent from text. A conversation element is received. An intent is determined by matching a domain independent relationship and a domain dependent term determined from the received conversation element to an intent included in an intent database that stores a plurality of intents and by inputting the matched intent into a trained classifier that computes a likelihood that the matched intent is the intent of the received conversation element. An action is determined based on the determined intent. A response to the received conversation element is generated based on the determined action and output.