Strengths4
- Comprehensive pipeline covering data acquisition, near-duplicate detection, clustering, lineage inference, and rigorous evaluation.
- Well-documented preprocessing and experimentation steps with clear instructions for running the frozen pipelines on new corpora.
- Use of appropriate and efficient algorithms (MinHash/LSH for duplicate detection, union-find for clustering) and statistically sound evaluation methods (bootstrapping, cluster-aware metrics).
- Pre-registered experiments and hypothesis-driven approach, enhancing the credibility of the findings.
