Hierarchical Intent Classification Model for User Session Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current flat-structured intent systems for user intent recognition are inaccurate due to independent intents without associations, leading to less accurate user intent determination in intelligent customer service agents.
Innovation Solution
A hierarchical intent classification model is employed, where user session text is processed to generate feature vectors, and probabilities are calculated using multi-level classifiers to assign user intents based on pre-trained models with hierarchical relationships among intent labels, reducing the need for extensive training data and memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a flat-structured intent system is used, then the system complexity is low, but the user intent recognition accuracy deteriorates
Solution Approach 1:
The patent segments the intent recognition system into a hierarchical structure with multiple levels (first-level intents, second-level intents, etc.). Each level divides the intent space further, creating a tree-like segmentation that organizes intents from general to specific. This segmentation allows the system to handle complex intent recognition by breaking it down into manageable hierarchical stages, improving accuracy without overwhelming system complexity.
Solution Approach 2:
The patent introduces a hierarchical dimension to the traditional flat intent system. Instead of treating all intents as equal-level categories in a single dimension, the system adds a vertical hierarchical dimension with multiple levels. This dimensional transformation allows intents to be organized in parent-child relationships, enabling more nuanced recognition while maintaining structured organization.
2Measurement precision
If extensive training data is used, then the user intent recognition accuracy is improved, but the memory occupation increases
Solution Approach 1:
The hierarchical intent structure segments the training data requirements across multiple levels. Each level of the hierarchy handles a subset of intents, distributing the training burden. This segmentation reduces the amount of training data needed at each individual level compared to training a single flat model on all intents simultaneously, thereby reducing overall memory occupation while maintaining recognition accuracy.
Solution Approach 2:
By introducing the hierarchical dimension, the system reduces the training data footprint in the horizontal dimension (number of intents at one level) by organizing them vertically across multiple levels. This dimensional reorganization allows the model to learn patterns at each hierarchical level with fewer examples, reducing total memory requirements while preserving or improving accuracy through the structured learning approach.
Data Source
AI summary
The disclosed embodiments describe methods, systems, and apparatuses for determining user intent. A method is disclosed comprising obtaining a session text of a user; calculating, by the processor, a feature vector based on the session text; determining probabilities that the session text belongs to a plurality of intent labels, the probabilities calculated using a multi-level hierarchal intent classification model, the intent labels assigned to levels in the multi-level hierarchal intent classification model; and assigning a user intent to the session text based on the probabilities.