A method of constructing knowledge graph for government procurement system. (17th March 2022)

• Record Type: Journal Article
• Authors: Zhang, Xiaochuan; Liu, Lu
• Is Part Of: International journal of wireless and mobile computing. Volume 21:Number 4(2021)
• Page Start: 332
• Page End: 341
• Publication Date: 2022-03-17
• Subjects: natural language processing -- knowledge graph -- knowledge extraction -- government procurement; Mobile computing -- Periodicals; Wireless communication systems -- Periodicals; 004.6
• Languages: English
• ISSNs: 1741-1084

Abstract

Knowledge graph can build a bridge between data and knowledge. Aiming at the problems of weak relevance and lack of knowledge of relevant business data generated by the operation of government procurement system for many years, resulting in difficult information sharing, slow service response and poor user experience, this paper proposes a knowledge graph construction method combining top-down method and bottom-up method. For the knowledge extraction model, this paper adopts a pipeline model which is different from the popular joint extraction model. Firstly, BiGRU-CRF is used to obtain the entities; secondly, the relational classification model BiGRU-Attention establishes the relationship between entities, even to form the minimum unit of knowledge graph. Finally, using our model, we achieve average F1 of 83.2 without using too much computing power. Our model can effectively construct a knowledge graph and improve the service quality in the system.