{"id":"W3121217868","doi":"10.1016/j.is.2021.101718","title":"Multi-label legal document classification: A deep learning-based approach with label-attention and domain-specific pre-training","year":2021,"lang":"en","type":"article","venue":"Information Systems","topic":"Text and Document Classification Technologies","field":"Computer Science","cited_by":73,"is_retracted":false,"has_abstract":false,"ca_institutions":"Thomson Reuters (Canada)","funders":"","keywords":"Computer science; Artificial intelligence; Multi-label classification; Domain (mathematical analysis); Training (meteorology); Machine learning; Training set","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001647781,0.001595032,0.001547157,0.00306448,0.001296409,0.002041968,0.003697452,0.002844192,0.005518426],"category_scores_gemma":[0.003075048,0.0006066195,0.00134016,0.002655275,0.0006670802,0.003144622,0.002437753,0.004677767,0.003903836],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001993077,"about_ca_system_score_gemma":0.002934932,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0187812,"about_ca_topic_score_gemma":0.03737109,"domain_scores_codex":[0.9987261,0.0001782531,0.00008935053,0.0004162437,0.000323773,0.000266426],"domain_scores_gemma":[0.9975689,0.0007699335,0.0001841106,0.0004338741,0.000845594,0.0001974988],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0003630729,0.001044306,0.002197383,0.0001791673,0.0001045106,0.0001294315,0.0001194567,0.01744452,0.01337623,0.003505846,0.03213159,0.9294044],"study_design_scores_gemma":[0.0000318825,0.00007231294,0.0007225749,0.00003942103,0.00005807085,0.00007363729,0.0000784345,0.976003,0.01143343,0.006745298,0.004716859,0.00002511691],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0714076,0.002294987,0.893616,0.002249743,0.0008290017,0.00051509,0.002856842,0.01746476,0.008766009],"genre_scores_gemma":[0.391906,0.0009645445,0.5704145,0.001121866,0.0005521409,0.0003537508,0.01052983,0.0006387031,0.02351882],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.0187812,"threshold_uncertainty_score":0.03734374,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03582749930198761,"score_gpt":0.2545681606306928,"score_spread":0.2187406613287052,"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."}}