Construction Data Processing With Cycle-Consistent AI Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing large-scale artificial intelligence systems for processing construction data are prone to generating non-factual outputs, or 'hallucinations', leading to misinformed decisions and potential catastrophic outcomes in precision-oriented and safety-related construction contexts.

Innovation Solution

A computer system and method that utilizes a first generator model to map unstructured construction data to structured entity relations and a second generator model to map structured entity relations to unstructured data, ensuring cycle consistency and preventing hallucinations through training with positive and negative reference data, and employing first-order logic and neural network reliability checks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If large-scale artificial intelligence systems are used to process construction data, then productivity and decision-making capability are improved, but the reliability deteriorates due to generation of non-factual outputs or hallucinations

Engineering Contradiction:
Improvedata processing capabilityVSAvoidfactual accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent introduces an intermediary verification mechanism between the AI system and the construction data. A validation module acts as a mediator that checks AI-generated outputs against the original construction data, specifications, and constraints before presenting them as final results. This intermediary layer prevents hallucinations from reaching the user while preserving the high productivity of the AI system.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements a feedback loop where the output of the AI model is continuously validated against the input construction data. When hallucinations or factual errors are detected, the system provides feedback to correct the generation process. This feedback mechanism ensures that the AI system maintains high productivity while improving reliability through iterative validation and correction.

Inventive Principle:
Principle #23Feedback

2Ease of operation

If AI models generate unstructured output data, then ease of operation and user interaction are improved, but the manufacturing precision deteriorates due to potential hallucinations and factual errors

Engineering Contradiction:
Improvequery response capabilityVSAvoidoutput accuracy
Core Design Contradiction:
Ease of operationVSManufacturing precision

Solution Approach 1:

The patent introduces an intermediary verification mechanism between the AI system and the construction data. A validation module acts as a mediator that checks AI-generated outputs against the original construction data, specifications, and constraints before presenting them as final results. This intermediary layer prevents hallucinations from reaching the user while preserving the high productivity of the AI system.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary validation and verification steps before finalizing the output. The AI model generates responses, but before presenting them to the user, the system pre-checks the outputs against the construction data repository, specifications, and logical constraints. This preliminary action ensures accuracy is maintained while preserving the natural language interface.

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If construction data is stored in unstructured format, then adaptability and data ingestion flexibility are improved, but the loss of information increases due to lack of structured entity relations

Engineering Contradiction:
Improvedata format flexibilityVSAvoidentity relation structure
Core Design Contradiction:
Adaptability or versatilityVSLoss of information

Solution Approach 1:

The patent implements a dynamic data representation system that can adapt between structured and unstructured formats based on the operational context. The system maintains construction data with explicit entity relations in a structured format for accurate processing, while allowing flexible querying and presentation in unstructured natural language when needed. This dynamic approach preserves information integrity while providing adaptability.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system employs a nested data structure where unstructured construction data contains embedded structured entity relations. Within the flexible unstructured format, key entities and their relationships are systematically organized and tagged, allowing the system to maintain adaptability for various data sources while preserving the structural information needed for accurate processing and analysis.

Inventive Principle:
Principle #7Nested doll (Nesting)

Data Source

PatentEP4641450A1System and method for processing construction data
Publication Date: 2025.10.29 BENETICS AG
  • EP4641450A1 patent drawingFigure 1~2
  • EP4641450A1 patent drawingFigure 3~4
  • EP4641450A1 patent drawingFigure 5~6

AI summary

A construction knowledge base comprises structured construction data, arranged in a knowledge graph (G) with entities and relations between the entities. A user query (X) is mapped to a set of entity relations (R̂) using a first generator model (Md), trained to map unstructured construction data to structured entity relations. A subgraph (Z) is retrieved from the knowledge graph (G), using the set of entity relations for the query (R̂). The subgraph (Z) is mapped to unstructured data (Y) as output for the query, using a second generator model (Mg), inverse to the first generator model (Md) and trained to map structured entity relations to unstructured construction data. The generator models (Mg, (Md)) are trained for cycle consistency, whereby structured entity relations output by the first generator model (Md) are input to the second generator model (Mg), and unstructured data output by the second generator model (Mg) is input to the first generator model (Md).