Ontology Label Sharing for Faster Machine Learning Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of supervised learning techniques is the scarcity and complexity of labeled input-output pairs, exacerbated by specialized subjects requiring expert knowledge, making it difficult for human trainers to effectively train machine learning models.
Innovation Solution
A method and system for sharing labels and associated information between ontologies to train machine learning models, allowing for efficient transfer of learning and reduced redundancy in memory storage, while maintaining confidentiality and leveraging expert resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If labeled training data is manually created by human trainers, then the quality and accuracy of training data improves, but the time and cost required for data preparation increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-processing unlabelled training data through automated labeling systems before the main model training process. Ontologies are constructed in advance, and unlabelled data is pre-tagged with potential labels based on ontology matching, so that when models need training data, high-quality labeled data is already prepared and ready for use.
Solution Approach 2:
The patent introduces ontologies as intermediary structures that mediate between unlabelled raw data and labeled training data. The ontology serves as a knowledge bridge that automatically generates labels for unlabelled data, reducing the need for manual human labeling while maintaining data quality through structured domain knowledge.
2Measurement precision
If expert trainers are used to label specialized data, then the accuracy of labels for complex subjects improves, but the availability and cost of training resources decreases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically generate and refine its own training data through ontology-based labeling and model-generated labels. The system serves itself by using its own resources (ontologies, unlabelled data pools) to create training data, reducing dependency on external expert trainers while maintaining label accuracy through iterative refinement processes.
Solution Approach 2:
Ontologies act as intermediary knowledge structures that encode expert knowledge in a reusable format. Instead of requiring continuous access to expert trainers, the ontology captures their expertise once and then automatically applies it to label large volumes of data, making expert knowledge scalable and continuously available without additional expert resources.
3Reliability
If multiple ontologies are maintained separately for different machine learning models, then model specialization and performance improve, but memory storage and data management complexity increases
Solution Approach 1:
The patent merges multiple ontologies into a unified ontology structure that serves multiple machine learning models. Instead of maintaining separate ontologies for each model, the system combines them into a shared ontology that can be leveraged by multiple models, reducing memory storage requirements and simplifying ontology management while maintaining model specialization through selective label usage.
Solution Approach 2:
The unified ontology is designed to be universal and multi-functional, serving as a shared knowledge base for multiple different machine learning models. The same ontology structure and labels can be applied across different models and tasks, making the ontology system more efficient and reducing the overall complexity of maintaining multiple separate ontology systems.
4Measurement precision
If extensive labeled training data is collected and stored, then model training accuracy improves, but the computational resources and storage requirements increase
Solution Approach 1:
The system performs preliminary action by pre-processing and organizing unlabelled data into ontology-structured formats with potential labels assigned in advance. This preliminary organization means that when models need training data, the data is already structured and ready for efficient processing, reducing the need to store and process large volumes of completely unstructured data.
Solution Approach 2:
Ontologies serve as intermediaries that transform raw unlabelled data into structured training data with meaningful labels. Instead of storing and processing vast amounts of unlabelled raw data, the ontology system creates a compact, structured representation that retains the essential information needed for model training while reducing storage requirements.
Data Source
AI summary
A method performed by a machine learning system that involves obtaining a first ontology that includes one or more labels. Each label is associated with a sample that includes text. The ML system is configured to use a particular label to retrieve one or more samples associated with the particular label. The method further involves receiving an identification of a label of a first ontology associated with a first machine learning model to share with a second ontology associated with a second machine learning model and sharing the label and the information with the second ontology. The method further involves training the second machine learning model using the shared information associated with the label.