Hierarchical Classification Model for Intent Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing classification models struggle with efficiently classifying user intents from natural language inputs due to the subtleties and ambiguity of language, especially when dealing with a large number of classes, leading to high computation power and memory usage, and often result in unsuccessful or time-consuming training processes.
Innovation Solution
A hierarchical classification model is employed, utilizing a tree structure with binary and multiclass classification models at multiple layers, where each node is associated with a classification model, allowing for efficient training and classification of inputs into a large number of classes by reducing the depth of the tree and using binary models for non-leaf nodes and multiclass models at leaf nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single classification model is used to classify inputs into a large number of classes, then the classification capability is improved, but the training time and memory usage increase significantly
Solution Approach 1:
The patent divides the single classification model into multiple binary classification models arranged in a tree structure. Each binary classification model classifies inputs into two groups, and this process repeats at each level until reaching leaf nodes with final class assignments. This segmentation reduces the complexity of training individual models and enables parallel training of multiple simpler models, significantly reducing overall training time while maintaining the ability to classify into many classes.
2Adaptability or versatility
If a single classification model is used to classify inputs into a large number of classes, then the classification capability is improved, but the memory usage increases significantly
Solution Approach 1:
The patent segments the large classification model into multiple smaller binary classification models distributed across a tree structure. Each binary model stores only the parameters needed for its specific classification task, rather than storing all class information in a single model. This segmentation dramatically reduces the memory footprint of each individual model while collectively maintaining the capability to classify into many classes through the hierarchical structure.
3Productivity
If a hierarchical classification model with multiple layers is used, then the training efficiency is improved, but the model complexity increases
Solution Approach 1:
The patent segments the classification task into multiple binary classification models organized in a tree structure with specific depth and branching factors. This segmentation transforms a single complex multiclass classification problem into multiple simpler binary classification problems that can be trained independently and in parallel, improving training efficiency while the structured tree organization manages the overall model complexity.
Data Source
AI summary
Techniques disclosed herein relate to generating a hierarchical classification model that includes a plurality of classification models. The hierarchical classification model is configured to classify an input into a class in a plurality of classes and includes a tree structure. The tree structure includes leaf nodes and non-leaf nodes. Each non-leaf node has two child nodes associated with two respective sets of classes in the plurality of classes, where a difference between numbers of classes in the two sets of classes is zero or one. Each leaf node is associated with at least two but fewer than a first threshold number of classes. Each of the leaf nodes and non-leaf nodes is associated with a classification model in the plurality of classification models of the hierarchical classification model. The classification model associated with each respective node in the tree structure can be trained independently.