Syntactic Analysis for Decision-making Support System in Engineering-Procurement-Construction (EPC) Field

Abstract

The Engineering-Procurement-Construction (EPC) field is one of the complex industries that span the entire project cycle from bidding to engineering, construction, operations and maintenance (O&M). However, most EPC companies are exposed to contract-related risks during bidding or project execution period due to lack of data-based systematic decision-making system within limited time. In particular, in the client-supplied bidding document (ITB) in the EPC project, the client tends to pass the risk to the contractor. Therefore, when the client is participating in the bidding phase of a project, to analyze the contract (ITB) within a limited time and detect the presence or severity of risk sentence or clauses is of utmost importance. To analyze and detect the risk clauses of the bidding documents, professional experience and knowledge of the bidding documents is required, and it takes a lot of time and efforts to analyze and respond to the bidding documents that require complex sentences and expertise. In this study, it was performed as a preliminary step toward building an engineering decision support system. When conducting the EPC project, the items that could be risky were conceptualized by converting into a data base, and the main risk syntax and were constructed for algorithm. Text information was extracted from the bidding document (ITB) using syntax matching and named entity recognition technology for risk extraction, allowing users to systematically analyze and make a clear decision. In this study, research team applied to AI technology in EPC risk analysis especially phrase matcher and named-entity recognition (NER). Critical Risk Check Which is rule-based algorithm using phrase matcher method automatically extracts converts toxin clauses into a database. This Module contains 4steps as unstructured data Standardization, Pre-processing, Risk Database, Matching Algorithms. Terms Frequency Module using NER Model and EPC risk data was created in a similar syntax and converted into a JSON file. This package module identifying the frequency and location of the entity in the contract. The NER techniques can extract similar phrases of risky keywords and phrases. Also, can be demonstrated with domain characteristics such as location, general proper nouns as a frequency Image visualization. Through the Modules to be provided on decision-making support system as a cloud service. For the future works, research team improve the decision-making support system to present risk standards and semantic verification package.