Hybrid AI Classifier for Taxonomies With Sparse Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI classifiers face challenges in accurately classifying data into taxonomies due to the lack of high-quality training data for certain categories and subcategories, impacting their accuracy.
Innovation Solution
A hybrid AI classifier that combines a trained machine learning (ML) model with a large language model (LLM) to classify data into taxonomies, leveraging the LLM for categories and subcategories with insufficient training data, using a combination of confidence scores to determine the most accurate classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a supervised machine learning model is used for classification, then classification accuracy is improved for categories with sufficient training data, but classification accuracy deteriorates for categories with insufficient training data
Solution Approach 1:
The patent combines a supervised machine learning model with a large language model (LLM) into a hybrid AI classifier. The supervised ML model handles categories with sufficient training data, while the LLM handles categories with insufficient training data by leveraging its pre-trained knowledge. This merging allows the system to maintain high classification accuracy across all categories, resolving the contradiction between accuracy for well-trained categories and reliability for poorly-trained categories.
Solution Approach 2:
The LLM acts as an intermediary for categories where the supervised ML model lacks sufficient training data. When the supervised model encounters categories with insufficient training data, the system transitions to using the LLM, which has been pre-trained on vast amounts of general knowledge. This intermediary approach ensures reliable classification even when specialized training data is unavailable.
2Measurement precision
If training data is collected for every category and subcategory, then classification accuracy is improved, but data collection time and resources increase
Solution Approach 1:
The LLM is pre-trained on vast amounts of general knowledge data before deployment. This preliminary action allows the LLM to handle classification tasks for categories with insufficient training data without requiring additional data collection. The pre-training establishes a foundation of knowledge that can be leveraged immediately, eliminating the need to collect training data for every possible category.
Solution Approach 2:
The LLM serves as a universal classifier that can handle multiple categories and subcategories across different domains without requiring category-specific training data. Its pre-trained knowledge base allows it to perform classification tasks universally, reducing the need for extensive data collection for each specific category while maintaining high accuracy.
3Ease of manufacture
If only supervised machine learning is used, then the system is simpler to implement, but the system cannot handle categories without training data
Solution Approach 1:
The system merges a supervised ML model (which is relatively simple to implement) with an LLM (which provides versatility). The supervised ML model handles the straightforward cases with sufficient training data, while the LLM provides the capability to handle categories without training data. This combination maintains implementation simplicity for the supervised portion while adding versatility through the LLM component.
Data Source
AI summary
System, methods, apparatuses, and computer program products are disclosed for generating and using a hybrid artificial intelligence classifier for classifying input into one or more nodes of a taxonomy. Training data is received for at least a first portion of the taxonomy and used to train a supervised machine learning (ML) model to classify input into the first portion of the taxonomy having training data. A large language model (LLM) taxonomy is determined for at least a second portion of the taxonomy. The hybrid AI classifier classifies input based on a first classification obtained by providing the input to the supervised ML, and a second classification obtained by providing at least the input and the LLM taxonomy to a pre-trained LLM.