Automatic labeling of river restoration project documents based on project objectives and restoration methods

River restoration projects are widely implemented around the world. Such projects are usually complicated endeavors. Engineering, environmental, and social impacts must be considered under various constraints. Gaining lessons from existing projects would be valuable in new project planning. However, because such projects vary greatly in objectives and used methods, it is difficult to pinpoint relevant projects for reference, even if a large number of cases are collected in a decision-support system. In this study, machine-learning-based approaches were used to classify the description documents for river restoration projects. Based on the characteristics of such projects, a deep neural network (DNN) was developed to label project objectives, and a dictionary-based multilabel classification (MLC) method was developed to label project methods. The resulting labels were validated using 1400 project description documents. In labeling project objectives, the method resulted in a weighted average f1-score of 0.82 and 0.55 for the training and testing dataset, respectively. In labeling project methods, the f1-score was found to be 0.70. Both results indicate that the developed automatic labeling methods perform satisfactorily. The labels attached to the project documents enable project planners to conveniently find the relevant documents for reference and understand the relationships among the objectives and methods.