{"id":"W3099008231","doi":"10.18653/v1/2020.findings-emnlp.129","title":"exBERT: Extending Pre-trained Models with Domain-specific Vocabulary Under Constrained Training Resources","year":2020,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":75,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"Air Force Research Laboratory; MediaTek","keywords":"Computer science; Vocabulary; Benchmark (surveying); Artificial intelligence; Embedding; Domain (mathematical analysis); Context (archaeology); Machine learning; Computation; Training (meteorology); Training set; Natural language processing; Language model; Algorithm; Mathematics","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":[0.001909953,0.002559124,0.001024876,0.001337817,0.0005620045,0.00128114,0.002704117,0.001644164,0.005444447],"category_scores_gemma":[0.006623025,0.001148963,0.001751395,0.001104499,0.0007697776,0.004164476,0.002706144,0.003958076,0.004342335],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009976545,"about_ca_system_score_gemma":0.00163157,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009831248,"about_ca_topic_score_gemma":0.02205742,"domain_scores_codex":[0.9992538,0.0002159283,0.00005070185,0.0002633539,0.0001252602,0.00009078321],"domain_scores_gemma":[0.9978544,0.001111196,0.00007205544,0.0005513905,0.0003171957,0.0000936618],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0004489058,0.0003791839,0.00530784,0.0005107268,0.0004626444,0.0004833716,0.0003756347,0.3830921,0.01235734,0.007935319,0.04839358,0.5402534],"study_design_scores_gemma":[0.00004739499,0.000123518,0.0006754639,0.00007008889,0.0000635672,0.0001368206,0.00006193091,0.9735346,0.004928887,0.01051032,0.009810834,0.0000365567],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.06176603,0.003031231,0.8935305,0.001222719,0.0006618623,0.0003465374,0.003393172,0.02857022,0.007477728],"genre_scores_gemma":[0.5163735,0.001902902,0.4268773,0.00216678,0.0005905197,0.0009635332,0.0257493,0.002763753,0.02261242],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.009831248,"threshold_uncertainty_score":0.01954806,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05795474199009502,"score_gpt":0.2343886152408255,"score_spread":0.1764338732507305,"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."}}