{"id":"W4410761886","doi":"10.1080/01615440.2025.2500309","title":"Unlocking the archives: Using large language models to transcribe handwritten historical documents","year":2025,"lang":"en","type":"article","venue":"Historical Methods A Journal of Quantitative and Interdisciplinary History","topic":"Handwritten Text Recognition Techniques","field":"Computer Science","cited_by":10,"is_retracted":false,"has_abstract":true,"ca_institutions":"Wilfrid Laurier University","funders":"","keywords":"Computer science; Natural language processing","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002033723,0.0002733782,0.0006515877,0.0008704535,0.0003878082,0.00005581835,0.00101462,0.00009758338,0.00001564011],"category_scores_gemma":[0.0002062196,0.0002022057,0.0003354488,0.0004924638,0.0001121645,0.0007956554,0.0006702095,0.0006455787,0.000001638834],"about_ca_system_candidate":true,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.004096563,"about_ca_system_score_gemma":0.0002257695,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004061153,"about_ca_topic_score_gemma":0.000008613775,"domain_scores_codex":[0.9968067,0.00114679,0.0009235339,0.0004158786,0.0003501145,0.0003570145],"domain_scores_gemma":[0.9977444,0.0009988808,0.000419245,0.0003846237,0.0002276989,0.0002250936],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.001058855,0.001272108,0.00005843403,0.0003012406,0.0006290931,0.0004318091,0.2215665,0.0001699489,0.0898625,0.1480619,0.1260279,0.4105597],"study_design_scores_gemma":[0.002780409,0.004437109,0.000184556,0.002230236,0.0005227504,0.0004516561,0.00810581,0.0769201,0.004571212,0.2243645,0.6739646,0.001467133],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.001532918,0.02578516,0.9647189,0.002922961,0.001678449,0.0002098213,0.000002119391,0.00005305369,0.003096593],"genre_scores_gemma":[0.03309767,0.0002019465,0.9629442,0.001153087,0.0001139248,0.00002276673,5.028678e-7,0.00002442783,0.002441475],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.5479367,"threshold_uncertainty_score":0.9997265,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08453474367525225,"score_gpt":0.4017216701108192,"score_spread":0.3171869264355669,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}