Text Analytics for Supporting Stakeholder Opinion Mining for Large-scale Highway Projects

For large-scale highway projects, late identification of stakeholder concerns often leads to design changes and duplication of effort, which may cause major project delays. This paper proposes a stakeholder opinion mining approach for helping transportation practitioners better identify the types of concerns in the early project stage. The proposed approach includes two major components: (1) stakeholder concern extraction, and (2) stakeholder concern classification. This paper focuses on presenting the proposed methodology and experimental results for stakeholder concern extraction, which extracts the words and phrases that describe stakeholder concerns from stakeholder comments on large-scale highway projects. In developing the proposed stakeholder concern extraction methodology, several supervised machine learning (ML) algorithms were tested and evaluated, and the effect of using a predefined name list as feature was also investigated. All the algorithms were tested on a testing data set of 200 comment sentences, which were selected from a comment collection including 1,849 stakeholder comments on five large-scale highway projects.