Transformer Tagging of Building Device Names From Unstructured Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Building configuration systems face challenges in configuring building devices due to the lack of standard naming conventions, making it difficult to automatically generate tags for building devices from unstructured object data.

Innovation Solution

A method and system that utilize machine-learning models, specifically transformer models with multi-head attention layers and tokenizers, to convert unstructured object names into structured tags by training on embeddings and position embeddings, enabling the generation of standardized names for building devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If machine-learning models are trained on unstructured object names to generate standardized tags, then automation and accuracy of tagging improve, but device and system complexity increases

Engineering Contradiction:

Improveautomation of tag generationVSAvoidcomplexity of machine-learning system

Core Design Contradiction:

Extent of automationVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary machine-learning model (specifically a transformer-based model with tokenizers and embedding layers) that mediates between unstructured object names and standardized tags. This intermediary system automatically performs the conversion task that would otherwise require manual intervention, resolving the contradiction by automating the tagging process while managing complexity through specialized AI architecture.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms the tagging problem by changing parameters from manual classification to machine-learning-based prediction. The system uses training data to learn patterns in unstructured names and automatically generates standardized tags, improving automation while the complexity is managed through parameter optimization during training rather than runtime complexity.

Inventive Principle:

Principle #35Parameter changes

2Productivity

If manual configuration methods are used for building devices, then system complexity remains low, but productivity and time consumption worsen

Engineering Contradiction:

Improvespeed of device configurationVSAvoidtime for manual tagging

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training the machine-learning model on extensive datasets of unstructured object names and their corresponding standardized tags. This preliminary training phase enables the system to automatically and quickly generate accurate tags during actual building device configuration, significantly improving productivity while reducing the time loss associated with manual tagging during deployment.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system enables self-service by allowing the machine-learning model to automatically generate standardized tags without human intervention. The model processes unstructured object names and produces standardized tags autonomously, eliminating the time-consuming manual configuration process and dramatically improving productivity in building device setup.

Inventive Principle:

Principle #25Self-service

3Measurement precision

If standard naming conventions are enforced during device installation, then configuration accuracy improves, but ease of operation deteriorates due to rigid requirements

Engineering Contradiction:

Improveaccuracy of object tagsVSAvoidflexibility in device naming

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The patent inverts the traditional approach by not requiring users to manually enforce standard naming conventions during device installation. Instead, the system accepts flexible unstructured names from users and automatically converts them to standardized tags through machine-learning, thereby maintaining ease of operation while achieving high accuracy in the final tags.

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The machine-learning model serves as an intermediary that bridges the gap between flexible user input (unstructured names) and rigid system requirements (standardized tags). This intermediary automatically performs the conversion, allowing users to operate with flexibility while the system receives accurately standardized data, resolving the contradiction between ease of operation and measurement precision.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20240273405A1Training and executing machine-learning models for building data conversion

Publication Date: 2024.08.15 JOHNSON CONTROLS TYCO IP HLDG LLP

AI summary

Systems and methods for training and executing machine-learning models for building data conversion are disclosed. A system can identify a plurality of unstructured object names each associated with a respective object tag. The plurality of unstructured object names can correspond to a plurality of building devices of a building. The system can determine a plurality of embeddings based on the plurality of unstructured object names. The plurality of embeddings can include a position embedding. The system can train a machine-learning model based on (i) the plurality of embeddings including the position embedding, and (ii) the respective object tag of the plurality of unstructured object names. The machine-learning model can be trained to generate corrected object tags for the plurality of building devices.