Current Research and Future Directions for Off-Site Construction through LangChain with a Large Language Model

Abstract

Off-site construction is well-known technology that facilitates parallel processes of manufacturing and construction processes. This method enhances productivity while reducing accident, cost, and environmental impact. Many studies have highlighted its benefits, prompting further encouragement of off-site construction. This study consolidates current research and charts future directions by reviewing the existing literature. However, reviewing papers is time-intensive and laborious. Consequently, generative AI models, particularly Large Language Models (LLMs), are increasingly employed for document summarization. Specifically, LangChain influences LLMs through chaining data, demonstrating notable potential for research paper reviews. This study aims to evaluate the well-documented advantages of off-site construction through LangChain integrated with an LLM. It follows a streamlined process from the collection of research papers to conducting network analysis, examining 47 papers to uncover that current research primarily demonstrates off-site construction’s superiority through cutting-edge technologies. Yet, a data deficiency remains a challenge. The findings demonstrate that LangChain can rapidly and effectively summarize research, making it a valuable tool for literature reviews. This study advocates the broader application of LangChain in reviewing research papers, emphasizing its potential to streamline the literature review process and provide clear insights into off-site construction’s evolving landscape.

1. Introduction

Off-site construction is among the most anticipated technologies globally, promising to reduce construction duration, as well as accidents at and the environmental impacts of construction sites [1]. The following reason explains why these advantages are possible. Simply put, off-site construction workflows are divided into manufacturing and construction processes. Components are manufactured during the former phase and assembled on-site during the latter [2]. This approach has accelerated significant advancements in the construction industry, with many studies identifying various advantages stemming from the application of off-site construction [2,3]. Major advantages include improving productivity, ensuring safer sites, reducing costs, and mitigating environmental emissions [1,4,5,6]. First, as previously mentioned, off-site construction simultaneously processes manufacturing and construction phases [2]. Productivity has long been recognized as the foremost advantage. This is achievable because assembly and construction occur simultaneously. Recently, this benefit was magnified by the scarcity of skilled workers [4]. Meanwhile, demand continues to rise in response to the need for rapid construction. Globally, as the demand for semiconductors increases, so does the need for factories to expedite their production [1,7]. To address this issue, off-site construction is used when building factories to meet the desired outcomes [1,7]. Thus, productivity stands as one of the most significant advantages of off-site construction [8]. Second, for many years, the construction industry has been recognized as more hazardous than others. Workers are exposed to risks such as heights, severe weather conditions, and the operation of various heavy construction machinery [9,10,11]. Therefore, many injuries and fatal accidents have occurred [12]. Several countries have enacted construction safety laws to reduce accidents at construction sites. According to a previous study, 90% of construction accidents occur during the construction phase [13]. Thus, off-site construction plays a significant role in mitigating accidents through technological advancements [14]. Actually, the number of workers at off-site construction sites is comparable to that at on-site construction sites [1]. However, in terms of man-days, off-site construction can see a reduction due to shorter working hours [1]. Thus, with the reduced probability of accidents, the number of accidents can also decrease [11]. Through this, off-site construction can create a safer environment than on-site construction sites [5]. Third, regarding cost, several studies have reported conflicting results [3,15]. Previously, some studies indicated that off-site construction was more costly due to materials and transportation expenses [16]. Currently, a growing body of research suggests that off-site construction is less expensive than on-site construction, attributing savings to lower labor wages, reduced machinery rental fees, and indirect costs due to shorter construction periods [12]. Lastly, with rising concerns about global warming, mitigating the environmental impact of the construction industry has become increasingly necessary [12,17,18]. The construction industry accounts for approximately 30% of environmental impact [1]. Thus, mitigating environmental impact has become a priority in the construction industry. To achieve this goal, off-site construction is regarded as an eco-friendly method. It reduces waste and energy usage of construction machinery during the construction phase [17]. These are significant advantages over on-site construction.

In summary, off-site construction offers various advantages over on-site construction, with technologies continuously evolving to enhance these benefits. As advancements progress, identifying current research and future directions is important. Review studies play a key role in this context, providing well-explained overviews and insights into state-of-the-art research areas [19]. Review studies on off-site construction have been conducted periodically. Li et al. (2014) analyzed 10 research journals focusing on prefabricated construction, aiming to assess the contributions and the institutions involved. They found that the United States was the most active country in this research area, with the majority of papers published in Automation in Construction [20]. Jang et al. (2021) explored off-site construction to identify research trends from a management perspective. Their findings indicated steady progress in research on construction periods, with safety topics comprising approximately 13% of all studies [21]. The research highlighted not only academic papers but also policies related to off-site construction. Luo et al. (2021) examined policies on prefabricated construction in China. Their review suggested that, despite various incentives, funding was minimal due to budgetary constraints on incentives [22].

Thus, review studies provide insights and future directions in specific areas. However, conducting literature reviews is time-consuming and labor-intensive, making it challenging to thoroughly investigate and study these topics [23]. Recently, to address this challenge, Large Language Models (LLMs) have been employed to facilitate review studies. Notably, LangChain is gaining prominence in advancing review research [24]. As a generative AI model, the LLM is trained on extensive text data [25]. To leverage the LLM, authors trained it with research papers to extract desired answers. Taiwo et al. (2024) explored the potential opportunities and challenges associated with applying generative AI in the construction industry [26]. Ghimire et al. (2024) investigated the opportunities and challenges of implementing generative AI models in the construction industry. They proposed using generative AI models as a framework for conducting literature research specific to the construction industry [27].

Despite the scarcity of LLM-utilizing review studies in construction, previous research has indicated their potential benefits and challenges. This study addresses two questions: (1) how effectively can an LLM summarize research papers and provide quick, clear answers? (2) can an LLM offer valuable insights to future readers through the research review process? The goal is to review literature on off-site construction, offering insights to future readers and to determine how an LLM can be optimized for review studies. For these reasons, this research aims to assess the current and future landscape of off-site construction, emphasizing its advantages by employing an LLM as a framework.

2. Materials and Methods

This research is structured into three phases: (1) collection of research papers; (2) organization of the LangChain model for review; and (3) the conducting of a network analysis (See Figure 1). Detailed explanations are provided in the following chapters.

2.1. Collection of Research Papers

As mentioned earlier, this study aims to provide insights into the current research and future directions of off-site construction by reviewing research focused on four key advantages: productivity, safety, cost, and environmental impact. To accomplish this, the study begins by collecting research papers from databases, with Google Scholar chosen for its comprehensive coverage across all subject areas and its ability to access a wide variety of research documents compared to other databases [28]. This study adheres to two conditions to ensure both quantitative and qualitative aspects are met.

First, regarding quantitative aspects, this study aims to gather a comprehensive collection of research papers. To identify relevant papers on off-site construction, we employed a search strategy that combined ‘target’ AND ‘advantage’ as keywords. ‘Target’ refers to off-site construction, with keywords such as [‘Off-site’, ‘Modular’, ‘Prefabricated’] or [‘Construction’, ‘Building’, ‘System’]. The ‘advantages’ pertain to ‘Productivity’, ‘Safety’, ‘Cost’, and ‘Environment’, leading to a defined search strategy with keywords like {‘Productivity’: [‘Productivity’, ‘Schedule’, ‘Period’, ‘Duration’], ‘Safety’: [‘Safety’, ‘Accident’, ‘Risk’], ‘Cost’: [‘Cost’, ‘Economic’, ‘Life Cycle Cost’], ‘Environment’: [‘Environment’, ‘CO₂’, ‘Life Cycle Assessment’]}.

Second, from a qualitative aspect, the review study seeks to provide valuable insights and guidance for future readers. To ensure the quality of the review, the following considerations are essential [22,29]. First, the research period should span 10 years to guarantee the most recent topics of study [22]. Second, this study includes research papers published in high-ranking journals in the field of off-site construction with a focus on those ranked within the Journal Citation Reports Q1 [29]. Furthermore, this study excludes similar review papers to ensure originality and avoid redundancy.

Through these considerations, this study gathered 47 research papers, which were initially read and reviewed manually. While there are no defined criteria of review studies, it is crucial to determine whether the research papers selected are sufficient. Taiwo et al. (2024) employed a method of screening research titles and abstracts, ultimately selecting 10 potentially relevant research papers [26]. Shooshtarian et al. (2020) conducted a review of outdoor thermal comfort in Australia, analyzing 25 research papers [10]. As previously mentioned, there are no established criteria for conducting review studies. Nonetheless, the selection of a sufficient number of research papers is essential. According to previous studies, selecting research papers based on appropriate standards and procedures can yield meaningful insights. Therefore, this study aims to assess the current state of off-site construction research and outline future directions, analyzing 47 research papers with the aid of a large language model (LLM). The collected research papers are listed in

Table 1. Overview of the research papers.

Advantage	Author (Year)	Title	Journal	Ref.
Environment	Dong and Ng (2015)	A life cycle assessment model for evaluating the environmental impacts of building construction in Hong Kong	Building and Environment	[17]
Environment	Dong et al. (2015)	Comparing carbon emissions of precast and cast-in-situ construction methods–A case study of high-rise private building	Construction and Building Materials	[30]
Cost	Mao et al. (2016)	Cost analysis for sustainable off-site construction based on a multiple-case study in China	Habitat International	[15]
Productivity	Arashpour et al. (2016)	Off-site construction optimization: Sequencing multiple job classes with time constraints	Automation in Construction	[31]
Productivity	Li et al. (2016)	Schedule risks in prefabrication housing production in Hong Kong: a social network analysis	Journal of Cleaner Production	[32]
Productivity +Cost +Environment	Jeong et al. (2017)	An integrated evaluation of productivity, cost and CO2 emission between prefabricated and conventional columns.	Journal of Cleaner Production	[33]
Productivity	Li et al. (2017)	Integrating RFID and BIM technologies for mitigating risks and improving schedule performance of prefabricated house construction	Journal of Cleaner Production	[34]
Cost	Hong et al. (2018)	Barriers to promoting prefabricated construction in China: A cost–benefit analysis	Journal of Cleaner Production	[35]
Productivity +Cost	Jang and Lee (2018)	Process, productivity, and economic analyses of BIM–based multi-trade prefabrication—A case study	Automation in Construction	[2]
Productivity +Cost + Environment	Navarro-Rubio et al. (2019)	Sustainability, prefabrication and building optimization under different durability and re-using scenarios: Potential of dry precast structural connections	Sustainable Cities and Society	[3]
Productivity	Goh and Goh (2019)	Lean production theory-based simulation of modular construction processes	Automation in Construction	[36]
Safety	Wu et al. (2019)	Perceptions towards risks involved in off-site construction in the integrated design & construction project delivery	Journal of Cleaner Production	[37]
Environment	Hao et al. (2020)	Carbon emission reduction in prefabrication construction during materialization stage: A BIM-based life-cycle assessment approach	Science of the Total Environment	[18]
Environment	Liu et al. (2020)	Cyber-physical system-based real-time monitoring and visualization of greenhouse gas emissions of prefabricated construction	Journal of Cleaner Production	[38]
Environment	Liu et al. (2020)	Real-time carbon emission monitoring in prefabricated construction	Automation in Construction	[39]
Productivity + Cost + Environment	Shahpari et al. (2020)	Assessing the productivity of prefabricated and in-situ construction systems using hybrid multi-criteria decision making method	Journal of Building Engineering	[40]
Productivity + Cost + Environment	Tavares et al. (2021)	Prefabricated versus conventional construction: Comparing life-cycle impacts of alternative structural materials	Journal of Building Engineering	[41]
Productivity	Martinez et al. (2021)	A vision-based approach for automatic progress tracking of floor paneling in off-site construction facilities	Automation in Construction	[42]
Productivity	Yazdani (2021)	Production scheduling of off-site prefabricated construction components considering sequence dependent due dates	Environmental Science and Pollution Research	[43]
Productivity +Safety +Cost	Lu and Zhu (2021)	Integrating hoisting efficiency into construction site layout plan model for prefabricated construction	Journal of Construction Engineering and Management	[44]
Environment	Tian and Spatari (2022)	Environmental life cycle evaluation of prefabricated residential construction in China	Journal of Building Engineering	[45]
Environment	Liu et al. (2022)	A dynamic simulation study on the sustainability of prefabricated buildings	Sustainable Cities and Society	[46]
Environment	Yuan et al. (2022)	How to promote the sustainable development of prefabricated residential buildings in China: A tripartite evolutionary game analysis	Journal of Cleaner Production	[47]
Safety	Mohandes et al. (2022)	Occupational Health and Safety in Modular Integrated Construction projects: The case of crane operations	Journal of Cleaner Production	[5]
Productivity	Shen et al. (2022)	Prefabricated construction process optimization based on rework risk	Journal of Construction Engineering and Management	[48]
Productivity	Jeong et al. (2022)	Learning-driven construction productivity prediction for prefabricated external insulation wall system	Automation in Construction	[4]
Cost + Environment	Jeong and Jeong (2023)	Quantitative methodology of environmental impact and economic assessment under equivalent conditions for prefabricated systems	Journal of Building Engineering	[1]
Cost + Environment	Kumi and Jeong (2023)	Optimization model for selecting optimal prefabricated column design considering environmental impacts and costs using genetic algorithm	Journal of Cleaner Production	[7]
Environment	Jayawardana et al. (2023)	A comparative life cycle assessment of prefabricated and traditional construction–A case of a developing country	Journal of Building Engineering	[49]
Safety	Zhang and Lin (2023)	Prediction of Ergonomic Risks and Impacts on Construction Schedule through Agent-Based Simulation	Journal of Construction Engineering and Management	[50]
Productivity	Liu et al. (2023)	Scheduling optimization for production of prefabricated components with parallel work of serial machines	Automation in Construction	[8]
Productivity	Xie et al. (2023)	A case-based reasoning approach for solving schedule delay problems in prefabricated construction projects	Automation in Construction	[51]
Cost + Environment	Pan and Zhang (2023)	Benchmarking the sustainability of concrete and steel modular construction for buildings in urban development	Sustainable Cities and Society	[52]
Productivity	Peiris et al. (2023)	Digitalising modular construction: Enhancement of off-site manufacturing productivity via a manufacturing execution & control (MEC) system	Computers & Industrial Engineering	[53]
Cost + Environment	Cheng et al. (2023)	Life cycle environmental and cost assessment of prefabricated components manufacture	Journal of Cleaner Production	[54]
Cost	Tatari (2023)	Simulating Cost Risks for Prefabricated Construction in Developing Countries Using Bayesian Networks	Journal of Construction Engineering and Management	[55]
Environment	Moghayedi and Awuzie (2023)	Towards a net-zero carbon economy: A sustainability performance assessment of innovative prefabricated construction methods for affordable housing in Southern Africa	Sustainable Cities and Society	[56]
Productivity	Hasan and Lu (2024)	Enhanced model tree for quantifying output variances due to random data sampling: Productivity prediction applications	Automation in Construction	[57]
Environment	He et al. (2024)	Evolutionary game analysis of prefabricated buildings adoption under carbon emission trading scheme	Building and Environment	[58]
Safety	Yao et al. (2024)	Exploring the intentional unsafe behavior of workers in prefabricated construction based on structural equation modeling	Environmental Science and Pollution Research	[59]
Cost + Environment	Ji et al. (2024)	Improving the performance of prefabricated houses through multi-objective optimization design	Journal of Building Engineering	[6]
Cost + Environment	Zhu et al. (2024)	Stackelberg game-based method towards carbon-economy equilibrium for the prefabricated construction supply planning: A case study from China	Sustainable Cities and Society	[60]
Environment	Du et al. (2024)	Dynamic simulation for carbon emission reduction effects of the prefabricated building supply chain under environmental policies	Sustainable Cities and Society	[61]
Environment	Gao et al. (2024)	Multi-information integration-based life cycle analysis of greenhouse gas emissions for prefabricated construction: A case study of Shenzhen	Environmental Impact Assessment Review	[62]
Environment	Huang et al. (2024)	Process-based evaluation of carbon emissions from the on-site construction of prefabricated steel structures: A case study of a multistory data center in China	Journal of Cleaner Production	[63]
Productivity + Cost + Environment	Yuan et al. (2024)	Simulation and optimization of prefabricated building construction considering multiple objectives and uncertain factors	Journal of Building Engineering	[64]
Productivity	Chen et al. (2024)	Vision-based real-time process monitoring and problem feedback for productivity-oriented analysis in off-site construction	Automation in Construction	[65]

, organized by year of publication.

In terms of the annual number of papers, at the start of 2015, only two research papers were published, indicating a lack of initial interest in the advantages of off-site construction. However, over time, the number of research papers has gradually increased. By 2023, 11 research papers had been published. In 2024, 10 research papers have already been published, and considering the remaining period for publishing this year, more research papers are expected. Although this may seem like a small number, Figure 2 and Figure 3 and

Table 2. Overview of the published journals by year.

	Automat Constr	Build Environ	Comput Ind Eng	Constr Build Mater	Environ Impact Asses	Environ Sci Pollut R	Habitat Int	J Build Eng	J Clean Prod	J Constr Eng M	Sci Total Environ	Sustain Cities Soc	Total
2015	0	1	0	1	0	0	0	0	0	0	0	0	2
2016	1	0	0	0	0	0	1	0	1	0	0	0	3
2017	0	0	0	0	0	0	0	0	2	0	0	0	2
2018	1	0	0	0	0	0	0	0	1	0	0	0	2
2019	1	0	0	0	0	0	0	0	1	0	0	1	3
2020	1	0	0	0	0	0	0	1	1	0	1	0	4
2021	1	0	0	0	0	1	0	1	0	1	0	0	4
2022	1	0	0	0	0	0	0	1	2	1	0	1	6
2023	2	0	1	0	0	0	0	2	2	2	0	2	11
2024	2	1	0	0	1	1	0	2	1	0	0	2	10
Sum	10	2	1	1	1	2	1	7	10	4	1	6	47

and

Table 3. Overview of the research advantages of off-site construction by year.

	Productivity	Safety	Cost	Environment	Productivity + Cost	Cost + Environment	Productivity + Safety + Cost	Productivity + Cost + Environment	Total
2015	0	0	0	2	0	0	0	0	2
2016	2	0	1	0	0	0	0	0	3
2017	1	0	0	0	0	0	0	1	2
2018	0	0	1	0	1	0	0	0	2
2019	1	1	0	0	0	0	0	1	3
2020	0	0	0	3	0	0	0	1	4
2021	2	0	0	0	0	0	1	1	4
2022	2	1	0	3	0	0	0	0	6
2023	3	1	1	2	0	4	0	0	11
2024	2	1	0	4	0	2	0	1	10
Sum	13	4	3	14	1	6	1	5	47

show a different perspective. The annual number of research papers indicates a significant upward trend, suggesting that research interest in off-site construction has been steadily growing (see the black line and the indigo dotted line labeled ‘Total’ in Figure 2 and Figure 3).

Referring to Table 1, research papers covering various areas have been published in different journals over the decade from 2015 to 2024. Although publications continue into 2024, this study includes them to cover the most current research focuses.

Upon reviewing these research papers, their characteristics can broadly be categorized into two groups according to the number of advantages they have: those with a single advantage and those with multiple advantages. Specifically, studies with one advantage often aim to integrate cutting-edge technology into off-site construction. For instance, Goh and Goh (2019) proposed an optimal workflow to enhance productivity, leveraging discrete event simulations derived from real factory-type construction site data and insights from expert interviews to boost industrial competitiveness [36]. Meanwhile, studies addressing more than two advantages often compare different construction methods, including on-site and off-site approaches, from various perspectives. Tavares et al. (2021) evaluated four types of structure (i.e., two off-site and two on-site components) in terms of construction duration, cost, and environmental impact from a life cycle perspective. Their findings indicated that prefabricated construction is faster, cheaper, and more eco-friendly, compared to conventional construction. Notably, among the prefabricated options, the lightweight steel structure outperformed the others [41].

For a detailed analysis of the collected research papers, the following two analyses were conducted.

2.1.1. Analysis of the Publication Journal

Table 2 and Figure 2 illustrate the distribution of collected research papers by journal on a yearly basis. The papers, ranging from a single paper to as many as 11, were published across 12 journals related to off-site construction, addressing four distinct advantages. Notably, about 40% of the papers, with 10 published in each, appeared in Automation in Construction (burgundy dotted line) and Journal of Cleaner Production (dark red dotted line). Publications in each journal showed a gradual increase over time. The focus of each journal varies: Automation in Construction primarily explores the application of information technology in the lifecycle of constructed facilities, whereas the Journal of Cleaner Production concentrates on waste prevention and enhancing efficiencies in energy, water, resource usage, and human capital.

Through the scope of these journals, this study can anticipate current trends. Presently, research in off-site construction is focused on adopting technology such as artificial intelligence, the Internet of Things, and computer-aided design, alongside efforts to minimize environmental impact. The application of these forms of technology varies based on the research advantages. For instance, Liu et al. (2020) investigated a real-time monitoring system for tracking environmental impacts during the construction phase, employing RFID, GPS, and various sensors. This led to the proposal of a novel data-collection methodology [39]. Second, a key goal within off-site construction research is achieving low-carbon construction sites by reducing waste and mitigating carbon emissions. Kumi and Jeong (2023) proposed an optimal arrangement of prefabricated columns by considering structural performance as a constraint variable, utilizing a genetic algorithm. They evaluated 53 column types, selecting those that best balanced CO₂ emission reduction with structural integrity [7].

2.1.2. Analysis of Research Advantages of Off-Site Construction

In general, approximately 70% of the off-site construction research focused on a single specific advantage (i.e., productivity (orange dotted line), safety (olive dotted line), cost (aqua dotted line), or environment (purple dotted line)). Seven studies investigated two advantages, while six studies examined three. Regarding the distribution, the number of papers exploring two or more advantages was relatively consistent.

As time progresses, the number of studies focusing on the four key advantages (productivity, safety, cost, and environment) is increasing. Among these, research predominantly targets specific advantages, notably productivity and environmental impact. The interest in productivity remains high, with a total of 20 papers published in this area, indicating its continued relevance. Conversely, there has been a recent surge in environmental impact research, with 25 papers published, most of which appeared in the last three years.

This indicates a growing interest in environmental impact issues. Regarding cost, research on this advantage has been less prevalent, suggesting that cost is often considered alongside other advantages rather than in isolation. This multi-faceted approach indicates that cost is seen as a potential disadvantage of off-site construction [16]. To address the complex relationship between cost and other factors, Jang and Lee (2018) undertook a comprehensive productivity and economic analysis on a case project involving multi-trade corridor racks at an exhibition complex. Their findings suggest that off-site construction projects need to reach a certain scale to achieve economic benefits. Such a scale enables worker training, enhances productivity, and ensures economic viability by reducing working hours [2].

Finally, interested in the safety aspect of research has notably increased since 2020. As previously mentioned, the construction industry is inherently hazardous, prompting the adoption of off-site construction to mitigate these risks. Research on safety has been approached from diverse perspectives. For instance, Yao et al. (2024) explored the theoretical models and factors influencing the intentional risk preference variable, and underscored the necessity for targeted strategies to address international unsafe behavior among construction workers in prefabricated construction [59].

2.2. Organization of LangChain Model for Review

LangChain is an open-source framework designed to facilitate the development of applications powered by LLMs. It comprises a Model I/O, retrieval, and various other modules. The goal of LangChain is to connect potent LLMs, like those from OpenAI, with a range of external data sources, thereby enabling the creation and enhancement of natural language processing applications [66]. The key concept of LangChain involves ‘chaining’, linking external sources to the operation of LLM prompts, enabling functionalities such as translation, summarization, and Q and A services [25,66].

In this study, LangChain is utilized to summarize research papers, providing insights into current research and future directions for off-site construction. The LangChain model is constructed as follows [25,66] (See Figure 1).

(1)

Data Collection and Extraction: This study collected 47 research papers in PDF format, representing unstructured data. These files were processed using a LangChain module, which facilitated the loading and parsing of the documents to prepare them for further analysis.

(2)

Splitting Chunks: In this phase, the processed data were divided into manageable chunks, which served as units for processing by the LLM. This chunking process was executed using LangChain’s text-processing modules, ensuring each chunk was optimally sized for input into the LLM.

(3)

Embedding: Each text chunk was then converted into a vector representation using OpenAI’s library. This step involved embedding the text into a high-dimensional space suitable for machine learning applications, which is crucial for enabling the LLM to effectively process and analyze the textual data.

(4)

Establishment of Vector Database: A vector database was created to store the embedded text chunks. This database was indexed using FAISS to facilitate the efficient retrieval of relevant text chunks based on semantic similarity. This indexing process ensured that the LLM could quickly access and utilize the most pertinent information for answering research questions.

(5)

Prompt Creation: To extract meaningful insights, prompts were generated to query the LLM. These prompts were designed to retrieve relevant chunks from the vector database and guide the LLM in generating coherent and comprehensive answers. The prompts focused on summarizing research findings and identifying trends and future directions in off-site construction.

(6)

Answer Generation: Using the retrieved chunks, the LLM, specifically ChatGPT 3.5-Turbo, generated answers to the posed questions. Users could specify the text type, length, and style for the answers. The generated answers were evaluated for clarity, relevance, and completeness. This process was repeated iteratively to ensure the robustness of the generated insights.

2.3. Conducting Network Analysis

Lastly, to gain deeper insights into the current research trends and relationships among the 47 collected research papers, network analysis was conducted, focusing on frequency of keywords to identify current research trends. This study employed VOSviewer 1.6.20, a software known for network analysis, for this purpose. Such visualization offers valuable insights into the relationships, patterns, and structures among the collected papers. The choice of visualization technique was tailored to the specific research objectives and characteristics of the data [67,68].

(1)

Keyword Extraction: The initial step involved extracting keywords from the 47 collected research papers. This extraction process was automated using text-mining techniques to ensure the comprehensive and accurate identification of relevant terms. Each keyword was then evaluated for its frequency of occurrence across the dataset. In this study, a total of 1508 keywords were extracted.

(2)

Threshold Setting: To enhance the clarity and relevance of the network visualization, a minimum frequency threshold for keyword inclusion was established. In this study, keywords needed to appear at least three times across the research papers to be included in the analysis. Setting the threshold below three made it difficult to identify the network effectively. This threshold helped filter out less significant terms and focus on the most impactful keywords.

(3)

Network Construction: Using VOSviewer, a co-occurrence network was constructed based on the extracted keywords. In this network, nodes represent keywords, and edges represent the co-occurrence of keywords within the same research papers. The strength of the connections indicates the frequency with which pairs of keywords co-occurred, highlighting the relationships between different research themes.

(4)

Clustering and Visualization: VOSviewer automatically grouped closely related keywords into clusters, with each cluster representing a distinct thematic area within the off-site construction research field. The visualization was enhanced with color-coding to differentiate between clusters, making it easier to interpret the thematic relationships and dominant research areas.

(5)

Interpretation of Results: The network visualization was analyzed to identify key themes and trends in the research. The most prominent keywords and their connections were examined to understand the focus areas of current research. For instance, the strong connection between keywords such as ‘prefabricated system’ and ‘environmental impact’ indicated a significant research interest in the environmental benefits of off-site construction.

3. Results and Discussions

This chapter presents insights generated by ChatGPT through the LangChain model, which was trained on 47 research papers. To garner valuable information, the study posed two questions. The first question sought to understand the ‘current research’ on off-site construction, examining five aspects: objectives, advantages, disadvantages, contributions, and limitations. The second question focused on the ‘future directions’ of off-site construction research, asking specifically about this area.

When querying the LangChain model, trained on 47 research papers, answers tended to focus on the most recently trained documents. To address this issue, the model was initially trained on each of the 47 research papers individually. Subsequently, 47 distinct answers were gathered and used to further train the LLM via LangChain, ensuring a broader representation of insights across all papers [24].

3.1. Current Research for Off-Site Construction in Terms of Four Advantages

In this chapter, LLM was asked to address five key aspects of the current state of off-site construction. The answers to each aspect are analyzed and presented sequentially, as detailed in

Table 4. Answers of current research for off-site construction.

	Diagnosis
Objectives	Current off-site construction research is comprehensive, covering production optimization, productivity enhancement, sustainable development, risk identification, and new technologies like Modular Integrated Construction (MiC) projects. It includes detailed analyses of case projects and literature reviews. Additionally, there is a focus on developing innovative platforms like RFID-enabled real-time BIM platforms to enhance construction management and schedule performance in prefabricated construction projects.
Advantages	Current research on off-site construction showcases a clear understanding of its numerous advantages, including enhanced productivity, quality control, reduced construction waste, and improved worker safety. Researchers have also delved into environmental benefits, energy efficiency, and potential carbon reduction targets. Overall, the research in this area seems advanced and comprehensive.
Disadvantages	Current off-site construction research is focused on identifying limitations in productivity analysis, problem recognition, and challenges in design, construction management, supply chain management, and assembly.
Contributions	The current research level in off-site construction is advanced, with a focus on productivity improvement, risk mitigation, stakeholder analysis, cost-benefit analysis, environmental impacts, carbon emissions monitoring, and comparison with traditional methods. Additionally, research explores innovative technologies like artificial intelligence, BIM, RFID, and CPS to enhance efficiency, co-ordination, and sustainability in off-site construction projects.
Limitations	Current research on off-site construction faces limitations like small sample sizes, limited geographic scope, and a lack of consideration for full life-cycle stages. It also relies on specific data sources, which may lead to inaccuracies in data collection and calculations. Moreover, there is a need for further development in addressing uncertainties, incorporating comprehensive data, exploring additional factors impacting construction efficiency and safety, and expanding the research scope beyond specific scenarios and industries.

.

Objectives: according to the LLM, the 47 research papers provided a comprehensive overview, addressing productivity, risk, and sustainability through innovative technologies such as RFID and BIM. Each paper presented a unique objective and approached problem-solving with a range of advanced technologies. For instance, Hong et al. (2018) investigated the development of cost-analytical frameworks, emphasizing the significance of cost-benefit analysis in China [35]. Dong and Ng (2015) conducted a detailed analysis of the environmental impact of precast concrete in Hong Kong, focusing on life cycle assessment. They demonstrated that the majority of environmental impacts are attributed to materials [17].

Advantages: for several decades, several studies have highlighted the benefits of off-site construction. The LLM identified key advantages as enhanced productivity, improved worker safety, and environmental benefits. Specifically, Jeong et al. (2017) compared prefabricated columns with steel-reinforced concrete (SRC) columns, finding that prefabricated columns could increase productivity by up to 40%. The authors argued that such improved productivity could significantly reduce construction time, offering high value at low cost [33].

Disadvantages: according to the answer, the most significant disadvantage is the challenge in design, particularly in relation to assembly. Navarro-Rubio et al. (2019) highlighted the difficulty of connecting columns to girders. Resolving this issue could lead to reductions in construction duration and environmental impact [3].

Contributions: the contributions of off-site construction largely mirror its advantages. To delve deeper, the most notable contribution is the demonstrated superiority of off-site construction across several advantages compared to on-site construction. Jeong and Jeong (2023) conducted a comparison between prefabricated and conventional components (column and girder), focusing on economic and environmental impacts under identical conditions. They noted that previous studies yielded varying results due to comparisons of different construction projects. By analyzing different components within the same construction project, the authors established the superiority of off-site construction in terms of both economic and environmental impacts [1].

Limitations: The current research on off-site construction faces limitations, notably the scarcity of data and a concentration on construction phases. The first limitation pertains to artificial intelligence; for the evaluation or prediction of off-site construction outcomes, a validated dataset is crucial to ensure reliable results. Hao et al. (2020) assessed the carbon emissions of a prefabricated office building from a life cycle perspective using BIM [18]. Similarly, Huang et al. (2024) provided insightful analysis on the environmental impact of off-site construction, focusing on a case study of a data center built with a steel structure [63]. Previous studies often concentrated on single case studies. While the number of case studies may vary according to the research objectives, a key reason for the reliance on small samples is likely due to the off-site construction market being smaller compared to that of on-site construction [69]. Previous studies have employed various methodologies to address this issue. Jeong et al. (2021) proposed learning-driven methods that generate datasets through well-known simulations. Given the challenges and time-consuming nature of data collection, the authors utilized simulations to produce datasets via iterative simulation runs [4].

In summary, current research efforts in off-site construction aim to demonstrate its superiority through the latest technologies (e.g., BIM, AI, and sensors), offering multiple advantages over traditional on-site construction, including improvements in productivity, safety, economy, and environmental impact. However, challenges remain, particularly in data collection due to limited case studies and issues related to field constructability. Solutions and future directions will be explored in the following chapter.

3.2. Future Directions of Off-Site Construction in Terms of Four Advantages

This chapter sought an LLM’s perspective on the future directions of off-site construction, as detailed in

Table 5. Answer of future directions for off-site construction.

Future Directions

Future research

Future research in off-site construction could prioritize developing decision criteria for assessing building technologies, addressing skill shortages in the construction industry, and exploring the environmental benefits of prefabrication. Comparing practices and outcomes of off-site construction in different countries is also crucial. Additionally, investigating the implementation of circular economy strategies, such as material reuse and recycling, to maximize the benefits of prefabricated construction, is essential. Collaboration between industry and academia, along with government intervention through funding and incentives, could help overcome economic challenges faced by developing countries.

. The answer is summarized as follows:

(1)

Developing decision criteria for assessing building technology. The first answer is related to the decision-making process. Sometimes, off-site construction has conflicting advantages between cost and environment. Amid these conflicts, a multi-criteria decision-analysis method is needed so that clients can intuitively choose off-site construction. Shahpari et al. (2020) proposed a hybrid multi-criteria decision-making method to pinpoint the most crucial factors for off-site construction, utilizing a questionnaire survey. This approach led them to identify management, planning, and cost as significant factors influencing productivity [40].

(2)

Comparative analyses of off-site construction across different countries indicate that developing nations are particularly sensitive to costs, often viewing off-site construction as less viable. Tatari (2023) explored cost risk factors in prefabricated construction, focusing on various developing countries. By reviewing relevant literature and conducting questionnaires, the study identified consistent cost risks associated with prefabrication in these regions. Key risk factors impeding off-site construction include ‘Machinery and technology’, ‘Direct costs’, and ‘Scheduling and planning’. Addressing these factors is crucial for promoting prefabricated construction in developing countries [46].

(3)

The implementation of circular economy strategies for material reuse and recycling is suggested to maximize sustainable off-site construction. Mitigating environmental impact is recognized as a key advantage of off-site construction, chiefly due to its ability to reduce waste through the repeated use of materials, underscoring its status as an eco-friendly technology [70]. Pan and Zhang (2023) examined two modular building projects in Hong Kong, comparing a concrete system with a steel system in the context of sustainable urban development. Their findings indicate that modular construction markedly surpasses traditional methods, achieving a 46–87% reduction in waste, and thereby alleviating the urban waste-disposal burden [52].

(4)

Collaboration among industry, academia, and government is essential to promote off-site construction in developing countries. Historically perceived as more expensive than on-site construction, off-site methods require incentives from governments to become more widespread. Jayawardana et al. (2023) explored the environmental benefits of prefabricated construction in developing nations, finding it could significantly reduce carbon emissions, especially during the construction phase. However, they noted limitations in addressing economic and social sustainability. Future research should focus on encouraging prefabrication adoption in regions with low socioeconomic status and overcoming the challenges of its nascent industry. Governmental support is key to raising awareness and fostering sustainable construction practices [49].

In summary, insights from the LLM indicate a shift towards foundational research to rejuvenate off-site construction, moving beyond mere technological advancements. Emphasis is placed on cost and environmental impact as pivotal areas for propelling off-site construction forward, rather than solely on productivity improvements. As depicted in Table 3 and Figure 3, future research in off-site construction is anticipated to intensify, with a particular focus on cost and environmental impact.

3.3. Network of Keyword Frequency Analysis of the Research Papers

Figure 4 showcases a network analysis conducted with VOSviewer, based on the 47 research papers collected. This analysis highlights the central themes in off-site construction according to four advantages, setting a minimum keyword frequency threshold of three. Out of 1508 keywords, 80 were identified as crucial to the network. The most notable keywords include ‘prefabricated system’ highlighted in blue, ‘worker’ in yellow, and ‘risk’ in turquoise, along with others like ‘environmental impact’, ‘benefit’, ‘efficiency’, ‘challenge’, and ‘government’. These keywords underscore the discussed advantages and future directions. Specifically, the blue cluster focuses on advantageous aspects related to the ‘prefabricated system’, illustrating the extensive research aimed at underscoring the benefits of off-site construction. Additionally, the red cluster revolves around ‘benefit’, ‘prefabricated building’, and ‘government’, indicating the necessity for governmental and industrial efforts to bolster off-site construction, with related terms such as ‘subsidy’, ‘market’, and ‘developer’ also featuring prominently.

3.4. Discussion

This study conducted an investigation into the current research and prospective pathways for off-site construction, with a focus on productivity, safety, cost, and environmental impact through a comprehensive literature review. These four advantages are widely recognized as the key benefits of off-site construction. Diverging from the traditional review methodologies, this research utilized an LLM to examine the current research and future directions of off-site construction. This innovative approach enabled the succinct summarization of 47 research papers, facilitating an in-depth analysis of existing studies and emerging trends.

The summary of the current research and future directions for off-site construction is as follows:

➓

Current Research: Off-site construction demonstrates superiority by utilizing the latest technologies (e.g., BIM, AI, and sensors), offering clear advantages over traditional on-site construction. Recent trends have indicated a growing interest in environmental impacts over productivity. A significant limitation remains the difficulty in data collection, attributed to the relatively few off-site construction projects. To address this, efforts have been made to enhance data collection and augmentation through various forms of technology.

➓

Future Directions: The focus is on selecting optimal solutions across multiple advantages and prioritizing sustainable technology over mere technological innovation.

At the beginning of this study, two key questions were posed to assess the efficacy of using a Large Language Model (LLM) for researching off-site construction:

(1)

How can an LLM summarize a research paper and answer questions quickly and clearly? This study formulated six questions regarding off-site construction, and utilized the OpenAI Python library for answers. The processing time per question, with hardware specifications of a CPU M2 Pro and 16 GB RAM, was under five seconds, indicating a rapid summarization capability. The LLM effectively condensed 47 research papers, despite their varying objectives, methods, results, and contributions. Each question was concisely answered in five sentences or fewer, neatly organizing the information while providing valuable insights for future readers.

(2)

Can an LLM offer future readers insights through the research review process? Assessing the depth of insight from the answers is challenging due to the lack of a clear definition of ‘insight’ in the context of literature reviews. However, the ability to generate well-organized research summaries, highlight contributions, and outline future directions suggests that LLMs hold significant potential for enhancing literature review processes.

These findings highlight the practicality and potential of LLMs in streamlining research reviews, offering clear, concise summaries and valuable insights into off-site construction.

While the LLM model proves highly beneficial for reviewing research papers, certain considerations remain. First, ensuring the reliability of paper summaries poses a challenge. Performance varies among LLMs, and the specific model utilized in this study, GPT-3.5 Turbo, exhibited a performance rate of approximately 60% [71]. To enhance performance, employing higher-performing LLMs is advisable. However, these more capable models come at a higher cost than the model used in this study, presenting a financial challenge. Alternatively, integrating other technologies with LLMs, such as few-shot learning, could offer a solution to this problem [72]. Second, the results demonstrated variability based on sequence learning. Initially, when training on all 47 research papers simultaneously, the outcomes were biased towards the last reviewed paper. This issue was addressed by individually training on each of the 47 papers and then retraining with the compiled results. Yet, to minimize costs, a more efficient method for collectively training on numerous research papers needs to be developed.

4. Conclusions

Off-site construction is widely recognized as a forward-looking technology within the construction industry, with many research papers published annually to advocate its adoption. This study aimed to conduct an investigation into the prominent advantages of off-site construction, such as productivity, safety, cost, and environmental impact to understand the current research and identify future directions. Until now, traditional literature reviews, however, have been time-intensive and laborious. To streamline this process, LangChain combined with a Large Language Model (LLM) offered a swift and efficient means to summarize research findings, providing targeted insights. Despite its potential, LangChain had yet to be extensively utilized for reviewing literature in this domain. Therefore, this research undertook an analysis of off-site construction, focusing on the aforementioned key areas, employing LangChain and an LLM to evaluate the present state of research and speculate on future developments.

The results suggest that current research on off-site construction focuses on employing the latest technologies to demonstrate its advantages, particularly in terms of productivity. However, there is a growing interest in the environmental impacts of off-site construction, a trend that is becoming more pronounced over time. Network analysis confirms this, showing the strongest connectivity between the prefabricated system and environmental impact. While there are still challenges to be addressed in reviewing research papers through LLMs, such as refining questioning methods and verifying results, the use of generative AI models for literature reviews is expected to become increasingly widespread.

This study makes contributions across technical, research, and social dimensions:

➓

Technical Aspect: Employing LangChain and an LLM facilitated the review of 47 research papers, streamlining what is traditionally a time-intensive and challenging process. This innovative approach, not yet widely adopted in construction research, enabled the clearer and more efficient summarization of the literature.

➓

Research Aspect: By analyzing papers on off-site construction from multiple aspects, this study provided a comprehensive overview of the current state of research and outlines future directions from six perspectives.

➓

Social Aspect: The findings underscore the importance of collaboration between the construction industry and government support to advance off-site construction. Recognizing that cost sensitivity is a significant concern for clients, the study highlighted the need for industry-led technological advancements and government incentives to mitigate cost risks and encourage the adoption of off-site construction.

This study acknowledges certain limitations. First, the focus was exclusively on the productivity, safety, cost, and environmental impacts of off-site construction, overlooking other potential benefits. Second, there was no comparative analysis of various LLMs, despite their diverse applications in research. Choosing the optimal LLM for literature review requires a balance between performance and cost, a comparison that presents challenges.

Future research will address these limitations by employing a multi-criteria analysis to select the most suitable LLM. With the chosen LLM, the scope of research on off-site construction will be broadened to encompass a wider range of topics.