Information requirement analysis for establishing intelligent natural language query interfaces in BIM-based construction projects
Emerging artificial intelligence (AI)-based natural language interface (NLI) systems show significant potential for enabling stakeholders to efficiently retrieve complicated building information models (BIM). Previous studies have shown many technical pathways, but they have not investigated which i...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Tsinghua University Press
2025-06-01
|
| Series: | Journal of Intelligent Construction |
| Subjects: | |
| Online Access: | https://www.sciopen.com/article/10.26599/JIC.2025.9180084 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Emerging artificial intelligence (AI)-based natural language interface (NLI) systems show significant potential for enabling stakeholders to efficiently retrieve complicated building information models (BIM). Previous studies have shown many technical pathways, but they have not investigated which information entities in complex BIM schemas and constraint types were most important for NLI-based data querying. This study investigates the information requirements for NLI-based BIM model retrieval. It begins with a survey of existing BIM query languages (BIMQLs) and software applications to identify popular information entities and constraints. We then recruited ten practitioners to create 200 queries and analyzed them to refine the information scope (IS) for NLI applications. Finally, we tested 14 selected queries via the NLI approach and other methods, revealing the types of queries that NLIs could better manage. This study identifies the most important information entities, constraint types, question forms, and condition combinations to develop intelligent NLI systems in BIM-based construction projects. The findings lay a crucial foundation for the advancement of AI-based NLIs by offering a definite IS, which can be used to generate training datasets or prompts for large language models. |
|---|---|
| ISSN: | 2958-3861 2958-2652 |