Infrastructure Ombudsman

Dataset

• The dataset can be found in the jsonl format in the dataset folder.
• It contains three columns: id, label, and source
• id indicates the unique comment id that can be used to get the text through the Reddit/YouToube API.
• label indicates the final label assigned to each of the text
• source indicates whether the comment/text is from Reddit or YouTube
• In case the full dataset with the text is required please email the corresponding author of the paper at mac9908@rit.edu

Code

• The code and configuration parameters for training the large language models LLAMA2 and MistralAI are in the llm.py file under the training folder.
• The usage and experimentations in their raw format are in the notebooks folder.

Citation

• If you use the dataset or code please cite the given paper:

@inproceedings{chowdhury2024infrastructure,
  title={Infrastructure Ombudsman: Mining Future Failure Concerns from Structural Disaster Response},
  author={Chowdhury, Md Towhidul Absar and Datta, Soumyajit and Sharma, Naveen and KhudaBukhsh, Ashiqur R.},
  booktitle={Proceedings of the ACM Web Conference 2024},
  year={2024}
}