Development of an AI-Powered System for Reviewing Construction Documents in Uzbekistan

39 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 RESEARCH ARTICLE DEVELOPMENT OF AN AI-POWERED SYSTEM FOR REVIEWING CONSTRUCTION DOCUMENTS IN UZBEKISTAN Dilmurod Rakhmatov1,* *Address correspondence to: rakhmato[email protected] Abstract The construction industry in Uzbekistan, like many others worldwide, faces significant challenges in document management and compliance with the vast and complex regulatory environment, notably "Construction Norms and Regulations" (CNR) and "State Standard of the Republic of Uzbekistan" (SSU) standards. This research aims to develop and evaluate an artificial intelligence (AI)-powered system specifically designed to automate the review of construction documents within the Uzbek context. Utilizing a combination of natural language processing (NLP) and machine learning (ML) techniques, the proposed system aims to significantly reduce the manual effort and time required for document review processes while improving accuracy and compliance rates. Our methodology encompasses the collection and annotation of a substantial corpus of construction documents, the development of an AI model trained on this dataset, and a rigorous evaluation of the system's performance against manually reviewed benchmarks. Results indicate a substantial improvement in both efficiency and accuracy of document review processes, with the AI system achieving 95% accuracy in compliance detection compared to 81% for traditional manual methods, and reducing review time from over 30 hours to under 4 hours per document set. The system demonstrated precision of 0.89, recall of 0.95, and an F1-score of 0.95 across diverse case studies in Tashkent, Samarkand, and Bukhara. The contributions of this study are twofold: first, it provides a novel application of AI technologies for automating document review processes in the construction industry of Uzbekistan, addressing specific regulatory requirements; second, it contributes to the broader field of construction informatics by demonstrating the potential of AI and ML technologies in enhancing regulatory compliance and efficiency. This research lays the groundwork for further exploration into AI-powered document management systems and their potential to transform the construction industry's approach to regulatory compliance and project management. Citation: Dilmurod Rakhmatov. Development of an AI-Powered System for Reviewing Construction Documents in Uzbekistan. Universa: International Journal of Research. 2025;1 Submitted 22 June 2025 Revised 29 July 2025 Accepted 27 September 2025 Published 2 November 2025 Copyright © 2025 Dilmurod Rakhmatov Exclusive licensee ThePublishGate. Distributed under a Creative Commons Attribution License (CC BY 4.0). Keywords AI in Construction, Document Review Automation, Natural Language Processing (NLP), Machine Learning, Regulatory Compliance, Construction Informatics, Uzbekistan Construction Standards, Document Management, Compliance Checking, Digital Transformation. 1 Lead AI Engineer, Ministry of Construction and Housing and Communal Services of the Republic of Uzbekistan, Tashkent, Uzbekistan * Lecturer, Tashkent State University of Economy, Tashkent, Uzbekistan 40 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 INTRODUCTION The advent of digital transformation initiatives within the construction sector of Uzbekistan heralds a paradigm shift towards more efficient, transparent, and accountable project management and execution. At the heart of this digital transition lies the critical process of construction document review, a procedural cornerstone that ensures projects not only comply with the national construction codes, namely “Construction Norms and Regulations” (CNR) and “State Standard of the Republic of Uzbekistan” (SSU) standards, but also adhere to the best practices in safety, environmental sustainability, and structural integrity. The significance of accurate and expedited document review cannot be overstated, as it directly influences project timelines, cost efficiency, and regulatory compliance, which are pivotal to the success and sustainability of construction endeavors [Toochukwu, 2025]. Historically, the construction industry in Uzbekistan, similar to its global counterparts, has grappled with the challenges posed by manual document review processes. These processes are not only time-consuming but are also prone to human errors, leading to inconsistencies in compliance and, at times, project delays. The manual review of construction documents presents opportunities for corruption, a risk exacerbated by the lack of transparency and the subjective discretion afforded to reviewers [Santos, 2025]. Such challenges underscore the urgent need for innovative solutions that can transcend the limitations of traditional review methodologies. The integration of AI into the construction sector's governance and management practices offers a promising avenue to address these pervasive challenges. AI, with its capabilities to process and analyze large volumes of data with unparalleled speed and accuracy, presents a transformative potential for automating the review of construction documents. Through the application of natural language processing (NLP) and machine learning (ML) algorithms, AI-powered systems can efficiently interpret, classify, and verify the compliance of construction documents against the established CNR and SSU standards, thereby significantly reducing the review time and minimizing human error [Adamu et al., 2014; Jakubik et al., 2022]. Despite the recognized potential, the deployment of AI in construction document review in Uzbekistan remains nascent, with limited research and development efforts focused on this area. The existing literature predominantly concentrates on the broader applications of AI in construction, offering generalized insights that, while valuable, do not specifically address the unique challenges and opportunities presented by the Uzbekistan construction sector [Leite et al., 2016; Kamaev, 2025]. This gap in the literature underscores a critical research void, necessitating focused investigation into the development, implementation, and evaluation of AI-powered systems tailored for the Uzbekistan construction industry's specific needs. The purpose of this study is to bridge this research gap by developing an AIpowered system designed to automate the review of construction documents in Uzbekistan. This system aims to leverage advanced AI/ML methodologies to enhance the efficiency, accuracy, and transparency of the document review process, thereby contributing to the broader objectives of digital transformation in the construction sector. Specifically, the study will explore the application of NLP techniques for the semantic analysis of textual content within construction documents and ML algorithms for pattern recognition and compliance verification. By doing so, it endeavors to offer a comprehensive solution that can mitigate the risks associated with manual review 41 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 processes, including inconsistencies in compliance, project delays, and corruption. Furthermore, this study will contribute to the existing body of knowledge by providing empirical evidence on the efficacy of AI-powered systems in improving construction document review processes. It will also offer practical insights that can inform policy formulation, regulatory practices, and project management strategies in the Uzbekistan construction sector and beyond. The integration of AI in construction governance represents a critical step forward in achieving digital transformation goals. By addressing the specific challenges of document review through innovative AI/ML applications, this study not only fills a significant research gap but also contributes to enhancing the efficiency, transparency, and accountability of construction projects in Uzbekistan. The subsequent sections will delve into the methodology employed in developing the AI-powered system, followed by an analysis of the system's performance and its implications for the construction industry's future. LITERATURE REVIEW The integration of AI in the construction sector signifies a monumental shift towards the automation of complex processes, including the review of construction documents. This literature review delves into the advancements in AI technologies within the construction sector, with a particular focus on the development of an AI-powered system for reviewing construction documents in Uzbekistan. The review is organized into thematic sections, each addressing a critical component of the system's development and application. AI in Construction Sector The global advancements in AI have paved the way for transformative changes in the construction industry, offering solutions for document classification, data analysis, and the integration of Building Information Modeling (BIM) with AI technologies. Document classification and management, a perennial challenge in the construction sector due to the voluminous and varied nature of documents, has seen significant improvements with the application of AI. Machine learning models have been adeptly employed to classify and manage documents, enhancing efficiency and reducing human error [Bafandegan Emroozi et al., 2024]. Furthermore, the integration of AI with BIM has facilitated more informed decision-making processes, optimizing project outcomes through data-driven insights [Panet al., 2023]. These advancements underscore the potential of AI to revolutionize traditional practices within the construction industry, setting the stage for the application of such technologies in Uzbekistan's construction sector. NLP for Document Understanding NLP stands at the forefront of enabling computers to understand, interpret, and manipulate human language. In the context of construction document review, NLP techniques such as PDF text extraction and Named Entity Recognition (NER) have been instrumental. PDF text extraction allows for the retrieval of textual data from documents, a fundamental step in processing and analyzing content. NER, on the other hand, identifies and categorizes key information, such as construction terms and compliance criteria, from unstructured text [Naik et al., 2023]. These NLP capabilities are crucial for automating the review of construction documents, enabling the identification and classification of pertinent information with respect to regulatory compliance. Computer Vision for Construction Drawings 42 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 The application of Computer Vision (CV) in the analysis of construction drawings introduces a novel approach to understanding graphical content in documents. CV techniques, combined with Optical Character Recognition (OCR), facilitate the recognition and interpretation of text within images and drawings, an essential aspect of construction documents. The development of OCR and CV pipelines has enabled the extraction of detailed information from construction drawings, supporting more accurate and comprehensive document reviews [Memonet et al., 2020]. This integration of CV and OCR technologies represents a significant advancement in automating the review of construction documents, particularly in recognizing layout features and annotations in drawings. AI Adoption in Public Sector The adoption of AI in the public sector, including e-government initiatives, highlights the broader implications of AI technologies for governance and public service delivery. Challenges related to governance, data privacy, and ethical considerations, however, remain pertinent issues. The implementation of AI systems within public sector governance, including the construction industry, requires careful consideration of these challenges to ensure transparency, accountability, and public trust [Agrawal, 2024]. The examination of e-government initiatives provides valuable insights into the potential and pitfalls of AI adoption in the public sector, informing the development of AI-powered systems for reviewing construction documents. Uzbekistan Context The context of Uzbekistan, characterized by its specific regulatory environment and digital transformation initiatives, presents both opportunities and challenges for the implementation of AI in construction document review. The CNR and standards (SSU) necessitate a tailored approach to the development of AI-powered systems, ensuring compliance with local regulations. The burgeoning digital platforms in Uzbekistan, alongside the push for e-government and digital governance, provide a conducive environment for the adoption of AI technologies in the construction sector [Gomboin et al., 2025]. However, the development and implementation of such systems must be attuned to the regulatory, cultural, and technological landscape of Uzbekistan. Research Gap Identification Despite the promising developments in AI technologies for construction document review, a specific investigation into their applicability within the Uzbekistan context remains limited. The integration of NLP and CV technologies for analyzing construction documents, while globally advancing, necessitates further research to adapt these technologies to the unique regulatory and linguistic nuances of Uzbekistan. Moreover, the effective adoption of AI in the public sector, particularly in governance-related aspects of the construction industry, requires a deeper understanding of the local governance framework and regulatory compliance mechanisms. This literature review highlights the need for a focused study on the development and evaluation of an AI-powered system for reviewing construction documents in Uzbekistan, addressing the specific challenges and opportunities presented by the national construction codes and standards. While global advancements in AI offer a robust foundation for automating construction document review processes, the adaptation and implementation of these technologies within the Uzbekistan context warrant further investigation. This research gap underscores the significance of developing a tailored AI-powered system that aligns 43 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 with the regulatory, linguistic, and cultural specificities of Uzbekistan, thereby enhancing the efficiency and accuracy of construction document reviews in the country. METHODOLOGY This section elucidates the comprehensive methodology adopted for the development of an AI-powered system designed to automate the review of construction documents in Uzbekistan. The system's architecture is meticulously outlined, encompassing data sources, preprocessing techniques, and the model pipeline. Subsequently, detailed descriptions of the NLP module, CV module, and the Rule-Based Compliance Engine are provided. Lastly, the model training and evaluation processes are thoroughly examined. The proposed system architecture integrates diverse data sources, including construction project documents, regulatory texts, and annotated datasets. These inputs undergo a preprocessing stage to standardize formats and cleanse data, thereby facilitating efficient data handling by subsequent AI modules [Panwar, 2024]. The architecture encompasses a multi-stage model pipeline, integrating an NLP module, a CV module, and a Rule-Based Compliance Engine to process, analyze, and evaluate the documents against Uzbek construction norms (CNR, SSU). Fig. 1 is showing the flow from data input through NLP/CV and Rule Engine modules to compliance engine output. FIGURE 1. AI Powered Document Review System Architecture. The NLP module is pivotal for document classification and information extraction, leveraging advanced techniques to interpret and categorize text-based content within the construction documents. Utilizing a combination of machine learning algorithms and LLMs, the module is adept at recognizing and extracting pertinent information, thereby facilitating an automated and accurate review process [Golilarz et al., 2024]. Fig. 2 44 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 illustrates the sequential processes involved in the NLP module, from raw text input to classified and extracted information output. The integration of LLMs is particularly beneficial for understanding the context and nuances of construction terminology and regulatory language specific to Uzbekistan. FIGURE 2. NLP Pipeline. FIGURE 3. Computer Vision Module for Construction Drawings. The CV module incorporates OCR, layout detection, and symbol recognition capabilities. This module is essential for processing scanned or photographed documents, converting images into machine-readable text, and identifying graphical elements and layouts indicative of specific document types or sections [Ravichandran et 45 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 al., 2025]. The symbol recognition function is particularly crucial for interpreting architectural drawings and construction plans, enabling the automated identification of standard symbols and notations used in the industry as depicted in Fig. 3. TABLE 1. System Technical Specifications Central to the system's functionality is the Rule-Based Compliance Engine, designed to automatically check the compliance of reviewed documents against the established Uzbek construction standards (CNR, SSU). This engine employs a comprehensive repository of rules derived from the regulatory texts, facilitating the automated evaluation of documents for adherence to national construction norms and standards. The development of this engine involved an extensive analysis of relevant legislation and standards to encode the rules accurately and effectively [Witt et al., 2024]. The model training phase involved the compilation of a dataset consisting of various construction documents, including project proposals, architectural plans, and compliance reports, annotated to facilitate supervised learning. The dataset was meticulously prepared, ensuring a representative mix of document types, layouts, and compliance scenarios to enhance the model's learning and generalization capabilities (Fig. 4). 46 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 FIGURE 4: Training Dataset Statistics. The evaluation of the model's performance was conducted using a validation set separate from the training dataset. Key metrics such as accuracy, precision, recall, and F1 score were employed to assess the model's effectiveness in document classification, information extraction, and compliance checking tasks. Preliminary results indicate promising levels of accuracy and efficiency, suggesting the potential of the AI-powered system to significantly enhance the construction document review process in Uzbekistan [Komilova & Tursunov, 2025]. The methodology outlined herein provides a robust foundation for the development and implementation of an AI-powered system tailored to the specific needs and regulatory environment of the Uzbekistan construction sector. The subsequent sections will delve into the results and discussion, highlighting the system's performance, potential implications for construction project management in Uzbekistan, and areas for future research. This objective and evidence-based approach ensures that the developed system is not only technologically advanced but also aligned with the cultural and regulatory specifics of Uzbekistan, thereby contributing to the digital transformation initiatives within the construction sector of the country. 47 Universa: International Journal of Research THEPUBLISHGATE JOURNAL SERIES Downloaded from https://thepublishgate.com/ Dilmurod Rakhmatov, 2025 | DOI: 10.5281/zenodo.17522501 RESULTS The development of an AI-powered system for reviewing construction documents in Uzbekistan has culminated in a series of empirical evaluations to assess the system's performance across various metrics. This section details the quantitative and qualitative results obtained from the deployment of the system, focusing on performance metrics, case study analyses, speed improvement comparisons, and accuracy in compliance detection. Furthermore, visual aids including confusion matrices, ROC curves, and sample outputs are employed to furnish a comprehensive understanding of the system's efficacy. The evaluation of the AI-powered system was conducted using a robust dataset of construction documents, including plans, permits, and reports, which were annotated based on the CNR and SSU standards. The system's performance was quantified using precision, recall, and F1-score metrics. Statistically significant improvements were noted in the system's ability to identify and classify relevant document features against the baseline models (Fig. 5). FIGURE 5: Model Performance Metrics. To further validate the system's practical utility, a series of case studies were conducted. These involved the review of a diverse set of construction documents from ongoing projects in Tashkent, Samarkand, and Bukhara. Each case study was designed to assess the system's capability to handle documents of varying complexity and in different formats. The AI system demonstrated a consistent accuracy rate of over 92% in identifying non-compliance issues when benchmarked against expert manual reviews. Notably, in a case involving the renovation of a historical site in Bukhara, the system adeptly identified discrepancies in the proposed materials that were not in alignment with the SSU standards for historical preservation. The accuracy of compliance detection is paramount for the system's adoption and trustworthiness. Through extensive testing, the system showcased an exemplary ability