AUTOMATING THE REVIEW OF CONTRACTUAL CORRESPONDENCE IN MEGA CONSTRUCTION PROJECTS USING TRANSFORMER-BASED MULTI-LABEL NLP

Date

2026-6-24

Author

Sağlar, Muhammet Ali

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

34
views

0
downloads

Cite This

Compliance with contractual requirements is a critical part of contract administration, particularly during the execution phase of mega construction projects. In such projects, clients and contractors exchange a substantial volume of formal correspondence concerning technical, contractual, financial, and administrative matters. The manual review of these documents requires considerable time and coordination among different departments, and important contractual implications may be overlooked when the correspondence volume exceeds the available review capacity. In response to this problem, this thesis proposes an automated correspondence review methodology based on natural language processing and machine learning techniques, with the objective of supporting the classification and interpretation of contractual communications. A dataset of 615 incoming correspondences was collected from six contractors on a multi-billion-dollar complex construction project. Documents were multi-labeled according to confidentiality status, involved departments, and specific contract clauses. This research evaluates single-task (ST) framework against a state-of-the-art multi-task (MT) learning framework utilizing a shared DeBERTa-v3 encoder. The training pipeline and different ML models integrate domain and task-adaptive pre-training, data augmentation, and bootstrap ensembling, along with specific hyperparameter and threshold optimizations. The proposed model achieved an 83% macro-F1 score on highly imbalanced, multi-label data. These results demonstrate that an automated, MT approach significantly enhances contract management during the project execution phase which is critical and prolonged stage of the project life cycle. By streamlining these workflows, project management teams can coordinate one of the daily activities more effectively, reliably, and with greater contractual precision

Subject Keywords

İnşaat Sözleşme Yönetimi

URI

https://hdl.handle.net/11511/119722

Collections

Graduate School of Natural and Applied Sciences, Thesis