Raman Similarity Metrics Benchmark: Spectra, Perturbations, and Evaluation Results

This dataset contains the data products used to evaluate Raman spectral similarity measures under controlled perturbations. The benchmark comprises 500 reference spectra and 50,000 perturbed comparisons, with 100 realizations per reference. The collection contains 249 public experimental or reference spectra from RRUFF, NASA RAMdb, the Raman Open Database, and OpenSpecy-derived sources, together with 251 synthetic spectra generated with RamanSPy. The archive includes per-record traceability and available experimental metadata, reference and perturbed Lorentzian peak tables, noisy-query fit results, perturbation descriptors, similarity scores, retrieval analyses, sensitivity and uncertainty summaries, article figures, and file-level SHA-256 checksums. The evaluated measures are Pearson correlation, cosine similarity, bounded-lag cross-correlation, cross-cosine, dynamic time warping, SFEC, and the peak-oriented similarity index S_J. The continuous noisy-query arrays are not stored as 50,000 separate files. The archive preserves the perturbed peak tables, applied perturbation descriptors, noisy-query fit tables, and complete metric table used to inspect or reconstruct the benchmark comparisons. The dataset supports the manuscript "Robust Comparison of Raman Spectra Under Various Perturbations: Benchmarking Common Similarity Metrics and a New Peak-Oriented Index." Per-record source identifiers, source URLs, checksums, and attribution guidance are provided in the traceability files and SOURCE_CITATIONS.md. Rights and reuse conditions for source spectra remain subject to the terms of their original providers.