LLM-Generated Intent Taxonomies for Dynamic User Behavior
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying user intent from log data is challenging due to its fluid nature and lack of context, especially in emerging modalities where user understanding, usage, and behaviors rapidly evolve.
Innovation Solution
The use of automated systems, specifically large language models (LLMs), to efficiently generate and validate intent taxonomies, allowing for accurate identification of user intents in data requests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If automated systems with large language models are used to generate intent taxonomies, then the accuracy and reliability of intent identification is improved, but the computing resource consumption increases
Solution Approach 1:
The system generates intent taxonomies in advance using large language models before actual intent identification tasks. By pre-computing and validating comprehensive intent categories and subcategories, the system prepares structured knowledge that can be efficiently queried during runtime, reducing the computational burden of real-time intent analysis while maintaining high accuracy.
Solution Approach 2:
The system creates a simplified representation of user intent through generated intent taxonomies that mirror actual user behavior patterns. These taxonomies serve as copycat structures that capture essential intent characteristics without requiring complex real-time analysis, allowing the system to use simpler, less resource-intensive methods for actual intent classification while maintaining reliability.
2Use of energy by moving object
If manual methods are used to create intent taxonomies, then computing resource consumption is reduced, but the time required for taxonomy generation and updates increases
Solution Approach 1:
The system replaces manual mechanical processes of intent taxonomy creation with automated large language models. Instead of requiring human analysts to systematically create and maintain intent categories, the system uses AI-generated content that automatically adapts to emerging user behaviors and modalities, dramatically reducing both time and computational resources required for taxonomy generation and updates.
Solution Approach 2:
The system dynamically adjusts taxonomy parameters by generating updated intent categories based on changing user behavior patterns and emerging modalities. Rather than requiring manual re-creation of entire taxonomies, the system modifies specific parameters and categories automatically, allowing rapid adaptation to new user intents while maintaining overall structural integrity and reducing time-to-update.
3Adaptability or versatility
If intent taxonomies are made comprehensive to capture all user intents, then the coverage and accuracy of intent identification is improved, but the complexity of the taxonomy structure increases
Solution Approach 1:
The system segments user intent into hierarchical categories and subcategories generated by large language models. This segmentation organizes complex intent patterns into manageable, nested structures where broad intent types are divided into more specific subcategories, making the comprehensive taxonomy more navigable and easier to process while maintaining complete coverage of user behaviors across different modalities.
Solution Approach 2:
The system creates a dynamic taxonomy structure that automatically adapts to emerging user behaviors and modalities. Rather than a static hierarchical structure, the intent categories and subcategories can evolve over time as user patterns change, allowing the taxonomy to maintain comprehensive coverage while adapting its structure to reflect current user needs and reducing complexity through automated reorganization.
Data Source
AI summary
Methods, computer systems, computer-storage media, and graphical user interfaces are provided for efficiently generating and using intent taxonomies. In embodiments, training data, including data requests for information, is obtained. Thereafter, a model prompt to be input into a large language model is generated. The model prompt includes an instruction to generate an intent taxonomy, an indication of the training data to use for generating the intent taxonomy, and a taxonomy attribute desired to be used as criteria to generate a quality intent taxonomy. An intent taxonomy that includes user intent classes is obtained as output from the large language model. The intent taxonomy is analyzed to determine whether the intent taxonomy is valid. When the intent taxonomy is determined as valid, the intent taxonomy is provided for use in identifying user intent, and when the intent taxonomy is determined as invalid, the intent taxonomy is refined.