Landscape Architecture Standards Intelligent Q&A System
1. Problems Addressed and Product Positioning
In the landscape architecture industry, practitioners often encounter difficulties in searching for standards: there are hundreds of national, industrial, and local standards in total. To verify issues such as "park facility configuration" and "permeable pavement requirements", they have to search through multiple documents, which is time-consuming and prone to omissions. General question-answering tools either provide vague answers or occasionally contain professional errors (e.g., confusing greening topsoil standards). When grassroots staff or design teams handle multiple projects, it is challenging to quickly locate scenario-specific standard
Figure 1 Qualitative Evaluation Results of Various Models
To address these pain points, we have developed a prototype of an Intelligent Question-Answering System for Landscape Architecture Standards based on large language models (GPT-3.5-turbo) and retrieval-augmented generation (RAG) technology. Rather than replacing human judgment, the system is positioned as a "standard query auxiliary tool"—it integrates 15,135 landscape architecture standard question-answer pairs (covering national, industrial, and local standards) and can return accurate results with annotated standard sources through natural language interaction. The system is tailored to three user groups: frontline designers (for quick access to specifications), design firms (to meet standard requirements across multiple projects), and industry regulatory authorities (for verifying standard compliance basis). It provides technical support for "streamlining processes and enhancing accuracy" in industry standard queries.
Figure 2 Overall Framework of the Landscape Architecture Standards Intelligent Q&A System
2. Practical Value and Performance
Test results show that the system has fulfilled core requirements:
For designers: When inquiring about questions such as "What facilities are required for a 20-hectare park?", there is no need to browse multiple documents like the Code for Landscape Architecture Engineering Projects. Answers with standard numbers can be obtained within seconds, saving 80% of the time compared to manual retrieval.
For design institutions: When responding to clients' inquiries about "whether the plan complies with local standards", the system can quickly verify compliance, avoiding proposal impacts caused by standard omissions.
For regulatory authorities: When verifying project compliance, the original standard texts corresponding to answers can be directly accessed, reducing verification costs.
The system outperforms most general-purpose models in terms of standard query accuracy and professionalism. For example, when answering "specifications for street tree planting", the accuracy rate reaches 92%, and all answers are accompanied by specific standard sources (e.g., GB51192-2016), enhancing information credibility.
Figure 3 Interactive Interface of the Landscape Architecture Standards Intelligent Q&A System
Currently, the system can stably process text-based standard queries, but there is still room for improvement in response speed optimization, multimodal queries (e.g., uploading drawings to query corresponding standards), and supplementation of niche local standards. If you are a landscape design enterprise looking to integrate standard query functions, a landscape software team intending to add professional Q&A modules, or a government affairs platform needing to connect with standard verification tools, we welcome you to contact us. We can customize exclusive knowledge bases (e.g., focusing on segmented scenarios such as park design and vertical greening) or provide interface adaptation for existing systems, making standard queries more efficient and helping translate academic achievements into industrial practices.
