{"record": {"aliases": ["BullingerDB: A Dataset for Handwritten Text Recognition and Writer Retrieval"], "artifacts": [{"id": "arxiv:2605.30235", "kind": "paper", "url": "https://arxiv.org/abs/2605.30235"}], "categories": ["General AI", "Information retrieval", "Vision & 3D", "cs.CV"], "description": {"en": "A large-scale historical-document benchmark for multilingual handwritten text recognition and time-aware writer retrieval."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.30235", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.30235", "title": "BullingerDB"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.30235", "modality": null, "name": "BullingerDB", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_bullingerdb-a-dataset-for-handwritten-text_99c03eb7", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "acd25617aa701114f836f0c756f4e2367fbc70e63527c08836f779c4d73a0861", "relation": "evaluates_only", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T19:46:21.348598Z", "source_url": "https://arxiv.org/abs/2605.30235"}, "publisher": null, "released": "2026-05-28", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.30235", "source_url": "https://arxiv.org/abs/2605.30235"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-30235", "source": "claire_radar", "source_benchmark_id": "2605.30235", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["BullingerDB: A Dataset for Handwritten Text Recognition and Writer Retrieval"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Its large historical corpus, ICDAR presentation, explicit metrics, and stated dataset availability provide a credible basis for near-term specialist attention.", "score": 6}, "capabilities": ["Information retrieval"], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T19:46:21.348598Z", "inputHash": "94a42c500c1b4e02eb4bc03ceb14e160c5599ca1c23ca0cf971dcadcd99cb2ff", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "BullingerDB: A Dataset for Handwritten Text Recognition and Writer Retrieval", "canonicalNameSource": "paper_title", "decisionReason": "The source formally names the benchmark, identifies two evaluation tasks, reports concrete CER, temporal nDCG, and mAP scoring, and states that the dataset is available through Zenodo.", "evaluationEvidence": {"reuse": "Dataset available via zenodo", "scoring": "TrOCR, the best performing model, achieves a CER of 9.1%. For writer retrieval, we introduce a temporal nDCG metric to assess time-aware retrieval. While temporally coherent retrieval is achievable, mAP (78.3%) scores indicate challenges", "task": "We evaluate BullingerDB on text recognition and writer retrieval."}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T19:46:21.348598Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A large-scale historical-document benchmark for multilingual handwritten text recognition and time-aware writer retrieval.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "We present BullingerDB, a large-scale benchmark dataset for historical document analysis based on the correspondence of Heinrich Bullinger (1504-1575)."}, "familyId": "bmf_fc5175a0c82f", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_bullingerdb-a-dataset-for-handwritten-text_99c03eb7", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2605.30235", "pdf": "https://arxiv.org/pdf/2605.30235", "project": null, "report": "https://arxiv.org/abs/2605.30235"}, "metrics": [], "motivation": "We present BullingerDB, a large-scale benchmark dataset for historical document analysis based on the correspondence of Heinrich Bullinger (1504-1575).", "name": "BullingerDB", "oneLine": "A large-scale historical-document benchmark for multilingual handwritten text recognition and time-aware writer retrieval.", "paperTitle": "BullingerDB: A Dataset for Handwritten Text Recognition and Writer Retrieval", "primaryDomain": "General AI", "publication": {"evidence": "Accepted for presentation at ICDAR2026. Dataset available via zenodo", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2605.30235", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "presentation at ICDAR2026", "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-05-28", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Marco Peer and Anna-Scius Bertrand and Patricia Scheurer and Andreas Fischer"], "categories": ["cs.CV"], "id": "2605.30235", "title": "BullingerDB: A Dataset for Handwritten Text Recognition and Writer Retrieval", "type": "arxiv", "url": "https://arxiv.org/abs/2605.30235"}, "sourceUpdatedAt": "2026-05-28T17:05:20Z", "topics": ["cs.CV"], "venueAttempts": [{"decisionRaw": "Accepted for presentation at ICDAR2026. Dataset available via zenodo", "evidence": [{"level": "author-claim", "observedAt": "2026-08-19T11:05:13.395391Z", "rawValue": "Accepted for presentation at ICDAR2026. Dataset available via zenodo", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2605.30235"}], "reviewStatus": "accepted", "venueName": "presentation at ICDAR2026"}], "whyItMatters": "It supports comparable evaluation over 20,898 pages and 499,222 text lines spanning 796 writers and six decades."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
