{"id":"W4283790221","doi":"10.1609/aaai.v36i10.21369","title":"Post-OCR Document Correction with Large Ensembles of Character Sequence-to-Sequence Models","year":2022,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Handwritten Text Recognition Techniques","field":"Computer Science","cited_by":16,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"Dalhousie University; Consejo de Investigación, Universidad Nacional de Salta; Consejo Nacional de Investigaciones Científicas y Técnicas; Compute Canada","keywords":"Sequence (biology); Computer science; Optical character recognition; Character (mathematics); Artificial intelligence; Code (set theory); Natural language processing; Task (project management); Process (computing); Pattern recognition (psychology); Speech recognition; Image (mathematics); Mathematics; Programming language","routes":{"ca_aff":true,"ca_fund":true,"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":[0.0008282074,0.0002664317,0.0003419752,0.0003126479,0.0003533755,0.0001650742,0.002180527,0.00005974948,0.0001553649],"category_scores_gemma":[0.0001459764,0.0002082407,0.0001163952,0.001081813,0.0001768095,0.0008207054,0.0007987957,0.0004212194,0.00002482264],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001499329,"about_ca_system_score_gemma":0.0001787232,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001428188,"about_ca_topic_score_gemma":0.00002007072,"domain_scores_codex":[0.9972713,0.00005113685,0.0006512344,0.0006342041,0.0009652023,0.0004269256],"domain_scores_gemma":[0.9974739,0.00008510643,0.0005972194,0.0004060964,0.001324033,0.0001136457],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0002375846,0.0003363712,0.00008354571,0.00004888025,0.00002610161,0.000002381752,0.003996985,0.0002664278,0.311632,0.6031708,0.0002252295,0.07997371],"study_design_scores_gemma":[0.00004204752,0.00112587,0.00005543833,0.0002425541,0.00001370766,0.0000355888,0.001281388,0.03746524,0.85989,0.09938884,0.0001606911,0.0002985824],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7295202,0.00001825,0.247659,0.01017467,0.001097838,0.002429725,0.0001287015,0.0005084481,0.008463114],"genre_scores_gemma":[0.9914952,0.00002214508,0.007042742,0.0008833368,0.00002791062,0.0002481004,0.000002574162,0.00001785279,0.0002601439],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.548258,"threshold_uncertainty_score":0.8491809,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07268029614619377,"score_gpt":0.2963970258287851,"score_spread":0.2237167296825914,"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."}}