As digital twins transition from concept to industry standards, the construction sector is witnessing a growing demand for enriched Building Information Models (BIM) that encapsulate extensive data about various facets of the built asset lifecycle. However, as technology rapidly advances, it brings forth a significant challenge: effectively navigating the vast and complex data within these digital models.
Results from our ongoing research survey underscore this issue, revealing a general dissatisfaction with the current BIM search experience, primarily due to its lack of intuitiveness. Our research project was initiated in direct response to this emerging challenge. The AECOO industry is currently facing a critical bottleneck: the retrieval of information from within BIM systems.
To address this problem, we delved into the latest advancements in Artificial Intelligence (AI), exploring state-of-the-art methods to enhance intuitive text-based search capabilities within BIM platforms.
According to our ongoing research survey on “The Current State of Search in BIM”, a significant majority, 82% of participants (academics and professionals), stated that the user experience with the current search functionality is not satisfying. While 69% of respondents indicated that the nature of BIM workflows necessitates complex queries, text-based search methods are, by a large margin, the least commonly utilized method for information retrieval, despite their potential for simplicity and efficiency.
Focus on keywords over semantics:
The lack of interest in using text-based search has its roots in the inability of the existing search mechanisms to handle the complexity and context-sensitive nature of queries within interrelated BIM data. The current text-search tools lack the semantic depth necessary to effectively navigate the intricate layers of BIM information.
Overemphasis on accuracy scores:
The second gap is the disproportionate focus on accuracy scores in research outcomes. While high accuracy is undoubtedly desirable, this singular concentration often overshadows practical constraints faced in the industry. Factors such as computational cost, response latency, and integration challenges are frequently relegated to the background. This imbalance leads to solutions that, while technically impressive, may not be practically viable or adaptable to real-world industry scenarios
To tackle these challenges, we leveraged state-of-the-art AI techniques for neural semantic search. The term “neural” denotes the application of sophisticated deep neural networks dedicated to comprehending human language. This approach is the driving force behind emerging AI-powered tools such as ChatGPT. However, instead of relying on general-purpose AI tools/services, our proposed system is custom-designed to meet the unique requirements of the AECOO industry.
To make our tailor-made system adaptable across various applications and disciplines, we’ve introduced a semantic structure rooted in the Industry Foundation Classes (IFC), which is mature and widely used open data standard for BIM. Founded on such a robust semantic structure, along with state-of-the-art deep learning and Natural Language Processing (NLP) techniques, we engineered an end-to-end system capable of understanding and manipulating user queries over BIM data. The system is uniquely designed to comprehend the user’s intent, decode syntactic and semantic structures of the user’s text input, extract corresponding BIM entities, and map them to instances within native IFC-formatted BIM files.
Democratizing the accessibility and usability of BIM: Enhances BIM data access invites diverse professional insights, breaking down technical barriers; Simplifies interaction with BIM models, making crucial project data accessible to a wider audience.
Broadened stakeholder engagement and inclusive decision-making: Extends the potential reach of BIM tools beyond their traditional user base of highly skilled technical users; architects to contractors can easily navigate BIM data, focusing on strategic decision-making rather than complex query formulation.
Streamlined industry practices: Simplifies workflows, reduces errors, facilitates planning, and fosters collaborative efforts across various AECOO disciplines, leading to optimized and faster operations.
Open standard compatibility: Aligns with industry trends towards interoperability and standardized data exchange, crucial for managing complex built assets.
Future-ready solution: Sets a new standard in the AECOO sector, offering a scalable, adaptable, and user-friendly approach to BIM-based innovations.
In summary, our project bridges the gap between professionals’ need for timely and easy access to the right data, and the complex and intricate structures of built asset data within BIM platforms. This initiative represents a significant step forward in making extensive and complex built asset data more accessible and comprehendible for professionals with diverse backgrounds across the AECOO industry.
The ultimate vision of this research is to elevate the machine’s comprehension of domain-specific technical language in the AECOO industry. By doing so, we aim to enable intuitive user interfaces and automate workflows, thereby transforming how professionals interact with and benefit from BIM technology. The implication of this initiative goes beyond enhancing search efficiency; its reach extends to redefining the interaction paradigm between industry professionals and digital models.
