SCLog Supplementary Material

SCLog - Universal Segmentation Tool for Log Parsing This is the full replication package for the paper "Towards Universal Segmentation for Log Parsing", ICPC 2026. If you use our artifact, please kindly cite out paper:@inproceedings{le2026towards, title={Towards Universal Segmentation for Log Parsing}, author={Le, Van-Hoang and Bianculli, Domenico and Nguyen, Huy-Trung}, booktitle={2026 IEEE/ACM 34th International Conference on Program Comprehension (ICPC)}, year={2026}, organization={IEEE/ACM} } Requirements* python 3.1+ Directory Structureanalyze: empirical analysis scripts and resultsdatasets: datasets used for evaluationevaluation: evaluation scriptslogparser: adopted log parsing techniques and common utilssclog: SCLog implementation Instructions1. Virtual environment and dependenciesWe provide a preconfigured virtual environment in the venv folder. To activate the environment, run: source venv/bin/activate Deactivate the environment by running: deactivate To create a new virtual environment and install dependencies, run: python3 -m venv {NAME_YOU_PREFER}source {NAME_YOU_PREFER}/bin/activate pip install -r requirements.txt2. Run benchmarksTo run the evaluation scripts, navigate to the `evaluation` folder and run: cd evaluationpython Drain_eval.py # Run Drain evaluation python AEL_eval.py # Run AEL evaluationpython LFA_eval.py # Run LFA evaluation python LogPPT_eval.py # Run LogPPT evaluation3. Main resultsThe main results are in the analyze folder:empirical_results: contains the results for empirical study on log segmentation.results: contains the results for SCLog evaluation against SOTA log parsers (RQ1, RQ2). 4. Docker usageIf you prefer running the parsers in a containerized environment, use the provided Docker setup. Note: To use Docker, ensure you have Docker and Docker Compose installed on your machine (first-time setup instructions can be found on the Docker website and the Docker Compose website).# Using docker-compose (recommended; builds if needed) docker-compose builddocker-compose run --rm sclog python AEL_eval.py docker-compose run --rm sclog python Drain_eval.pydocker-compose run --rm sclog python LFA_eval.py docker-compose run --rm sclog python LogPPT_eval.py # Pass SCLog flags for ablations as needed, e.g.: docker-compose run --rm sclog python AEL_eval.py -w -c -s 5. Ablation study To evaluate the contribution of each component (or a combination of components), you can specify the hyperparameters when running the evaluation scripts `{PARSER}_eval.py` in the `evaluation` folder. The hyperparameters are:-w (SCLogW): use whitespace for segmenting log messages.-c (SCLogC): do not use context-aware segmentation.-s (SCLogS): do not use similarity-based merging.-d (SCLogD): use specified delimiters for segmentation.-wp (SCLogWP): do not use regex preprocessing. Example commands:- Local: python AEL_eval.py -w -c -s- Docker Compose: docker-compose run --rm sclog python AEL_eval.py -w -c -s