The transportation project decision making process has long been criticized for resulting in frequent delays for the development of important projects. For large-scale transportation projects, late identification of stakeholder concerns have been identified as a major contributor to the delay. In order to help transportation decision makers better identify stakeholder concerns in the early project stage, this paper proposes a methodology for stakeholder opinion classification, which classifies the opinion expressions extracted from stakeholder comments on large-scale transportation projects into different concern groups. In developing the proposed methodology, both supervised methods and unsupervised methods were evaluated and compared. For the supervised methods, several machine learning algorithms were tested. For the unsupervised methods, different word-embedding methods were investigated. All methods were trained on a dataset of 1,369 comment sentences, and were tested on a dataset of 376 comment sentences. Both, the training and testing datasets were randomly selected from a comment collection for eight large-scale transportation projects nationwide. This paper focuses on presenting and discussing the proposed stakeholder opinion classification methodology and the experimental results.