Hierarchical Intent Discovery for Chatbot Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing chatbot technologies face challenges in accurately identifying the intent of user-problems when dealing with families of products or services, especially when customer queries can relate to multiple products or use varying natural language expressions.
Innovation Solution
A context-based multi-granularity intent discovery method using a hierarchical framework with a machine learning model, trained on a dataset of user-problems, to establish intent levels through classification or clustering algorithms, focusing on key intent features and improving intent detection quality by removing noise and considering context.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a chatbot uses a single-level intent classification system, then the system complexity is low, but the intent detection accuracy deteriorates when dealing with families of products or services
Solution Approach 1:
The patent divides the intent detection system into multiple hierarchical levels: first-level intents represent broad product or service categories, while second-level intents represent more specific user needs or actions. This segmentation allows the chatbot to progressively narrow down the user's intent, improving detection accuracy for complex product families without requiring a single overly complex classification model.
Solution Approach 2:
The patent introduces a hierarchical dimension to the intent classification system, transforming a flat single-level classification into a multi-level hierarchical structure. This dimensional change enables the system to capture both broad and specific intents systematically, improving accuracy for diverse product families while maintaining manageable complexity at each level.
2Adaptability or versatility
If the chatbot considers all possible product terms and expressions, then the coverage of user queries is complete, but the noise in training data increases reducing detection quality
Solution Approach 1:
The hierarchical intent structure segments user queries into progressively more specific categories. At each level, the system focuses on a subset of relevant terms and expressions, rather than considering all possible terms simultaneously. This segmentation maintains comprehensive query coverage while reducing noise by concentrating on context-relevant features at each hierarchical level.
Solution Approach 2:
The patent applies local quality by tailoring the classification focus to each hierarchical level. First-level intents use broader terms appropriate for product categories, while second-level intents use more specific terms appropriate for user actions or needs. This localized approach to term selection improves detection quality by using the most relevant vocabulary at each stage without introducing unnecessary noise from unrelated terms.
3Measurement precision
If the chatbot removes context information to focus on new features at each granularity level, then the detection of new content improves, but the loss of contextual information reduces overall intent understanding
Solution Approach 1:
The hierarchical structure performs preliminary classification at each level before moving to the next. First-level intents establish the broad context and filter relevant user queries, then second-level intents focus on specific details within that context. This preliminary action at each stage improves new content detection while preserving overall context understanding through the hierarchical progression.
Solution Approach 2:
The patent implements a nested hierarchical structure where second-level intents are nested within first-level intents. Each level contains and builds upon the context established by previous levels. This nesting allows the system to focus on new content at each level while maintaining awareness of the broader context through the hierarchical containment relationship.
Data Source
AI summary
Context-based multi-granularity intent discovery, in relation to determining the intent of a user-problem associated with one of a plurality of products, includes identifying each of the plurality of products in a base intent level of a hierarchical intent data structure, creating a first granularity level below the base intent level, and creating a subsequent granularity level. The first granularity level includes groupings of the plurality of user-problems created according to a first common intent feature; the common intent feature is related to a respective one of the plurality of products. The subsequent granularity level includes sub-groupings of the user-problems of a respective grouping with each sub-grouping created according to a second common intent feature. The second common intent feature of each sub-grouping is determined without consideration of the first common intent feature.