Sentiment analysis model for public construction projects using KoBERT

Abstract

The development of real-time communication technology has accelerated and enlarged conflict propagation, resulting in social conflicts in modern society. Social conflict significantly impacts the execution of a public construction project. Conflict of interest between stakeholders may cause construction delays and even cease the project. Against this backdrop, the authors aim to develop a proactive conflict management system based on Natural Language Processing (NLP) for public construction projects. As a point of departure, this paper introduces a social acceptance assessment model using Social Network Service (SNS) data. The authors employ Korean Bidirectional Encoder Representations from Transformers (KoBERT) for the text model development. The proposed model has two main functions. One is to filter out irrelevant text with the execution of a construction project. The other is to measure the degree of agreement on an ongoing construction project as social acceptability considering the semantic context of input text. Then, the keyword analysis is done to draw out the main thesis of social conflicts in public construction projects. An illustrative case application is done to validate the model’s applicability and to illustrate how sentiment analysis and keyword analysis be done. The proposed model is expected to detect potential social conflict signals through real-time monitoring, enabling the proactive reaction to prevent social conflicts during the execution of public construction projects.