NLU Concept System Domain Adaptation via Ensemble Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current natural language understanding (NLU) frameworks face challenges in achieving domain specificity due to the use of generic word vector distribution models, leading to poor performance in specific domains.
Innovation Solution
Implementing a concept system within the NLU framework that applies a concept model to extract intents and other information from user utterances, enhancing domain specificity by using sample utterances from an intent-entity model specific to the client's domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a generic word vector distribution model is used in the NLU framework, then the system can process a wide variety of domains, but the NLU performance precision deteriorates in specific domains
Solution Approach 1:
The patent applies local quality by training a domain-specific concept model using concept search data from a particular domain, thereby making that specific part of the NLU system optimized for local domain requirements while the overall system maintains generic capabilities through the ensemble scoring architecture
Solution Approach 2:
The NLU framework is segmented into multiple independent scoring systems including a generic word vector distribution model and a domain-specific concept model. Each segment handles different aspects of NLU, allowing the system to maintain versatility at the generic level while achieving precision through domain-specific segmentation
2Measurement precision
If a domain-specific concept model is implemented, then the NLU performance precision improves in specific domains, but the device complexity increases
Solution Approach 1:
An ensemble scoring system acts as an intermediary that coordinates between the generic word vector distribution model and the domain-specific concept model. This mediator manages the complexity by orchestrating multiple scoring mechanisms without requiring complete system redesign
Solution Approach 2:
The concept search data structure and processing pipeline are designed with universal applicability across different domains. The same framework infrastructure handles both generic and domain-specific concept extraction, reducing overall system complexity through multi-functional design
Data Source
AI summary
A natural language understanding (NLU) framework includes a concept system that performs concept matching of user utterances. The concept system generates a concept cluster model from sample utterances of an intent-entity model, and then trains a machine learning (ML) concept model based on the concept cluster model. Once trained, the concept model receives semantic vectors representing potential concepts extracted from utterances, and provides concept indicators to an ensemble scoring system. These concept indicators include indications of which concepts of the concept model that matched to the potential concepts, which intents of the intent-entity model are related to these concepts, and concept-relationship scores indicating a strength and/or uniqueness of the relationship between each concept-intent combination. Based on these concept-related indicators, the ensemble scoring system may determine and apply an ensemble scoring adjustment when determining an ensemble artifact score for each of the artifacts extracted from an utterance.