VizBrick: A GUI-based Interactive Tool for Authoring Semantic Metadata for Building Datasets

SangKeun LeeOSTI 2023 · OpenAlex

Abstract

The Brick ontology provides a unified semantic metadata schema to standardize the description of physical, logical, and virtual building assets. While Brick models enable significant benefits in data interoperability and automation, their creation typically requires specialized knowledge of semantic technologies like RDF and Turtle syntax. This "usability gap" often forces domain experts into a slow, manual cycle of trial and error. We present VizBrick, a graphical user interface (GUI) tool designed to assist users in creating Brick models interactively. VizBrick simplifies the process through keyword searches, TF-IDF-based concept suggestions, and rule-based relationship mapping. This paper demonstrates VizBrick's utility in transforming real-world building datasets into machine-readable semantic models without requiring deep expertise in ontology declarations.

Introduction

Smart building research relies heavily on high-quality data, yet building datasets are notoriously heterogeneous. Sensors and systems often use non-standardized naming conventions, making cross-building automation difficult. The Brick ontology addresses this by providing a standard vocabulary for building assets and their relationships. However, the barrier to entry remains high for building scientists who understand the physical layout but lack the computer science background to write RDF code. VizBrick was developed to "lower the floor" for these experts, providing a visual environment to bridge the gap between raw data and standardized semantic models.

The VizBrick Workflow

VizBrick introduces a streamlined, three-step workflow that guides the user from raw metadata to a finalized Brick model.

1. Metadata Review

Users begin by importing a standard CSV file containing data labels and descriptions. The tool provides a dedicated interface to clean and modify this metadata, ensuring a solid foundation before the mapping process begins.

2. Interactive Entity and Relationship Creation

To accelerate the mapping process, VizBrick leverages Natural Language Processing (NLP). Instead of manually browsing thousands of Brick classes, the tool uses a TF-IDF (Term Frequency-Inverse Document Frequency) engine to analyze data labels and suggest the most likely matching Brick concepts.

VizBrick's suggestion engine uses TF-IDF to map metadata labels to Brick classes with confidence scores.

As shown in Figure 5, the tool provides a ranking list and scores for potential matches (e.g., mapping "auxHeat1" to a "Run Time Sensor"), allowing users to quickly validate and apply mappings.

3. Rule-Based Mapping and Export

For large-scale datasets, manual mapping is unfeasible. VizBrick includes a Rule-based Edge Creator that allows users to define string-pattern rules to generate relationships in bulk. For example, a user can define a rule that automatically links any node containing the string "room" to a specific "HVAC Zone" entity using the isPartOf relationship.

The Rule-based Edge Creator allows users to define patterns for batch relationship generation, significantly improving scalability.

Once the model is complete, VizBrick provides a real-time graph visualization. The final model can be exported in the standard RDF Turtle (.ttl) format, ensuring full compatibility with the broader Brick software ecosystem.

Key Insights and Results

VizBrick demonstrates that domain expertise can be successfully decoupled from technical ontology syntax. By applying NLP techniques like TF-IDF to non-standard sensor names, the tool recognizes the linguistic clues inherent in building data. Furthermore, the rule-based approach addresses the scalability challenges of high-density commercial buildings.

The tool's effectiveness was validated using the Ecobee Donate Your Data (DYD) dataset, successfully transforming thermostat data from 1,000 homes into valid Brick models. VizBrick is released as open-source software, registered with the US Department of Energy’s Office of Scientific and Technical Information (OSTI).

Limitations and Future Work

While VizBrick significantly improves usability, the accuracy of its suggestion engine remains dependent on the quality of the input metadata. Future work will focus on improving UI scalability for extremely large buildings with tens of thousands of data points and integrating real-time semantic validation to ensure that user-defined relationships do not violate the internal logic of the Brick schema.

Summary updated

Abstract

Brick ontology is a unified semantic metadata schema to address the stand-ardization problem of buildings' physical, logical, and virtual assets and the relationships between them. Creating a Brick model for a building dataset means that the dataset's contents are semantically described using the standard terms defined in the Brick ontology. It will enable the benefits of data standardization, without having to recollect or reorganize the data and opens the possibility of automation leveraging the machine readability of the semantic metadata. The problem is that authoring Brick models for building datasets often requires knowledge of semantic technology (e.g., on-tology declarations and RDF syntax) and leads to repeated manual trial and error processes, which can be time-consuming and challenging to do with-out an interactive visual representation of the data. We developed VizBrick, a tool with a graphical user interface that can assist users in creating Brick models visually and interactively without having to understand the Re-source Description Framework (RDF) syntax. VizBrick provides handy ca-pabilities such as keyword search for easy find of relevant brick concepts and relations to their data columns and automatic suggestions of concept mapping. In this demonstration, we present a use-case of VizBrick to show-case how a Brick model can be created for a real-world building dataset.
BibTeX
@inproceedings{lee2023vizbrick,
  author = {SangKeun Lee},
  title = {{VizBrick: A GUI-based Interactive Tool for Authoring Semantic Metadata for Building Datasets}},
  booktitle = {OSTI},
  year = {2023},
}
Figure