During the implementation stage,parties of construction projects often need to retroact the liability subject based on the contract swiftly and accurately to avoid disputes.However,lengthy and complex construction contracts make this manual process time-consuming and error-prone.To tackle the problem,this paper presents a high-quality dataset through custom-built preprocessing based on the 2017 FIDIC silver book and fine-tunes the BERT-base model under the deep learning framework to increase the accuracy of the sentence-with-liability classifier.In addition,a rule-based liability-subject extractor and fuzzy retrieval algorithm are combined to establish a hybrid NLP model that automates the liability retroact approach using the FIDIC silver book as a basic.The proposed model outperforms previous models with regard to high efficiency,flexible retrieval,and robust application.The empirical study further underpins the effectiveness and effectiveness of the proposed model for contractual liability retroaction for international EPC projects.
维普
万方数据
