Postwar Japanese Social Movement Documents: OCR/HTR and Lexical Retrieval Evaluation Dataset

This dataset provides a reproducible benchmark for evaluating lexical retrieval from OCR/HTR transcriptions of postwar Japanese student and social movement materials. Version 1.0.0 contains a 106-page reference corpus, 69 exact-match queries, 475 positive query-page relevance judgments, page-level recognition outputs for 12 experimental conditions, and a standalone Python script reproducing the Stage 3 retrieval metrics. The benchmark was initially developed for the JINMONCOM 2026 study and can also serve as a common evaluation basis for subsequent OCR/HTR research. The reference transcriptions follow the upstream dataset v1.1.0 (DOI: 10.5281/zenodo.21798820), and trained OCR/HTR models are separately available (DOI: 10.5281/zenodo.23291038). This release preserves the fixed evaluation data used in the original study without revising the underlying transcriptions. Future releases may extend the evaluation conditions or benchmark data, with changes documented separately for reproducibility.