| Title |
Development of a RAG-Based System for Supporting Construction Arbitration Responses - Focusing on Construction Arbitration between Project Owners and Contractors - |
| Authors |
Chanhee Lee ; Han Soo Kim |
| DOI |
https://dx.doi.org/10.6106/KJCEM.2026.27.1.017 |
| Keywords |
Construction Disputes; Construction Arbitration; Response Support; RAG; LLM |
| Abstract |
Construction projects involve multiple stakeholders, which increases the likelihood of disputes. In particular, a construction company’s initial response in arbitration can significantly influence the outcome. Analyzing past similar cases is essential for formulating an effective response strategy. However, such analyses are still largely manual and often rely heavily on legal experts, resulting in substantial time and cost burdens. Recently, there have been efforts to automate this process using large language models (LLMs), but these face limitations such as hallucinations. Retrieval-Augmented Generation (RAG) offers a way to address these limitations by combining the internal knowledge base of LLMs with external knowledge sources. The objective of this study is to develop a response support system based on construction arbitration rulings using RAG, thereby assisting construction companies in formulating effective strategies during arbitration. To this end, the system was developed through a seven-stage process, and its performance was evaluated using two scenario-based tests. Based on these tests, the system's accuracy and consistency were assessed through LLM-based evaluation, yielding reliable results with scores above 4.0 on a five-point Likert scale. This study demonstrates that the proposed system can serve as a useful tool for construction companies frequently involved in dispute situations when establishing their response strategies. |