Processed and clustered UDRP decisions from 2015-2025

Instructions to use the artifacts Organization Our processed dataset is under data/full-dataset.jsonl.gz The generated clusters and statistics are under data/clusters The templates folder contains the template of prompt we used to process the data. code contains the source codes. Installation Create a Python virtual environment and activate it Install the packages in requirements.txt 1. Download disputes content in Markdown Compile a dataset of processed domain disputes like those you can generate using our previous tool. The file must be gzipped. Run the script download_disputes.py and pass as argument the file with the URLs and pass an outfile. You can try running python3 code/download_disputes.py -f data/sample-disputes.jsonl.gz -o sample-downloaded-disputes.jsonl.gz to execute the code on a random sample of 20 disputes. The file is saved in gzip format 2. Extraction of facts, panel conclusion, and summmarization of the arguments Set up the key for your Google AI Studio by following the instructions here. Configure the key in your environment by following the instructions for your operating system. Unix-system users can add export GEMINI_API_KEY="<KEY>" to their bash setting file. Run the script extract_facts_and_conclusions.py, from the root directory. It expects three arguments: the input file, which should be under the format created by download_disputes above; an output file, saved as a gzipped-JSONL file; and a folder to which the files created during the request to the LLM are dumped. The output is saved to the file you passed as argument. 3. Clustering of the summaries Set up the key for Voyage by following the instructions here. Run the script cluster_summaries by passing a file with the same format as the one generated in the previous step. You also need to indicate the folder to which the clusters, and performance metrics should be saved. 4. Analyzing the clusters. Execute the notebook evaluate_clusters.ipynb. It will produce a figure similar to Figure 9 in the appendix.