AI Semantic Model Generation for Ontology Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional semantic modeling approaches face challenges such as repetitive manual tasks, difficulty in addressing diverse stakeholder needs, accessibility and integration of proprietary data schemas, unification of data formats, and lack of effective interaction and interconversion between different formalisms, leading to inefficiencies and incomplete data integration.
Innovation Solution
Utilizing Large Language Models (LLM) and Large Multimodal Models (LMM) to automatically generate and enhance semantic models by determining and incorporating missing data elements, adapting to context, and presenting models in user-friendly interfaces, thereby addressing the limitations of traditional tools.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual semantic modeling is performed using traditional tools (whiteboards, Draw.io, Protégé), then domain experts can create conceptual models, but the process becomes repetitive, time-consuming, and error-prone due to manual mapping and transformation between different formalisms
Solution Approach 1:
The patent introduces an intermediary system comprising a data model generator, data model aligner, and code generator that automatically transforms conceptual models into aligned data models and executable code. This intermediary automation layer eliminates manual repetitive tasks while preserving model accuracy, directly resolving the contradiction between modeling precision and time investment
Solution Approach 2:
The system performs preliminary alignment of data models with upper ontologies and standards during the generation phase rather than requiring manual alignment later. By pre-aligning data models with established ontologies (such as ECLASS) and generating standardized code structures upfront, the system eliminates subsequent manual adjustment work while ensuring model accuracy from the start
2Adaptability or versatility
If traditional semantic modeling tools are used, then basic data modeling is possible, but addressing diverse stakeholder needs and maintaining multiple abstraction levels in sync becomes difficult
Solution Approach 1:
The patent creates a universal data model structure that serves multiple stakeholders and purposes simultaneously. The system generates a core data model that can be aligned with different upper ontologies for different domains (e.g., ECLASS for construction, other ontologies for different industries) and produces code in multiple languages. This multi-functional approach allows a single modeling effort to satisfy diverse stakeholder requirements without increasing management complexity
Solution Approach 2:
The system segments the semantic modeling process into distinct modular components: conceptual model creation, automatic data model generation, ontology alignment, and code generation. Each module handles a specific aspect of stakeholder requirements independently. This segmentation allows different abstraction levels and stakeholder perspectives to be managed separately and integrated systematically, reducing overall complexity
3Reliability
If manual alignment to upper ontologies is performed, then semantic standards can be maintained, but the process requires significant manual effort and iteration
Solution Approach 1:
The patent replaces the mechanical manual process of ontology alignment with an automated computational system. The data model aligner automatically maps generated data models to upper ontologies using algorithmic matching and transformation rules, substituting human manual alignment work with automated mechanical processes. This maintains semantic standard compliance through systematic rule-based alignment while dramatically increasing development speed
Solution Approach 2:
The system performs self-service ontology alignment by automatically generating alignment mappings between the data model and upper ontologies without requiring manual expert intervention. The alignment process is embedded within the automated generation workflow, allowing the system to self-correct and self-align models against semantic standards continuously, maintaining reliability while enabling rapid iterative development
4Loss of information
If proprietary data schemas are captured manually, then organizational knowledge can be preserved, but integration into standard ontologies becomes difficult and time-consuming
Solution Approach 1:
The patent introduces an intermediary alignment layer that bridges proprietary data schemas and standard ontologies. The system captures proprietary organizational knowledge in the data model generation phase, then automatically aligns this proprietary data with standard upper ontologies through computational mapping. This intermediary alignment process preserves proprietary information while enabling seamless integration with standard ontologies, eliminating the difficulty of manual integration
Data Source
Figure 1
Figure 2
AI summary
A semantic model shall be completed or corrected in a comfortable way. Therefore, there is provided a method of generating a semantic model of a technical process or technical unit including the step of preparing (S1) a data model structure of the technical process or technical unit. An artificial intelligence unit determines (S2) a data element, whereby the data model structure is input into the artificial intelligence unit. The semantic model is generated by incorporating the data element into the data model structure.