{"id":"W4396882454","doi":"10.48550/arxiv.2405.06563","title":"What Can Natural Language Processing Do for Peer Review?","year":2024,"lang":"en","type":"preprint","venue":"TUbilio (Technical University of Darmstadt)","topic":"Topic Modeling","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"Queen's University; Okanagan University College; University of British Columbia, Okanagan Campus; University of British Columbia","funders":"Office of Naval Research; Bundesministerium für Bildung und Forschung; Deutsche Forschungsgemeinschaft; Natural Sciences and Engineering Research Council of Canada; European Commission; Australian Government; Alberta Machine Intelligence Institute; Canadian Institute for Advanced Research; National Science Foundation","keywords":"Operationalization; Computer science; Pace; Process (computing); Peer review; Artificial intelligence; Technical peer review; Field (mathematics); Aside; Data science; Political science; Linguistics","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0009271246,0.0002969025,0.000609382,0.0002013203,0.0001097231,0.0002444792,0.002426316,0.0003017154,0.00002029435],"category_scores_gemma":[0.0001965107,0.000303001,0.0004021182,0.0003400445,0.0001389704,0.0003521616,0.004193004,0.001134445,0.00000971032],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002664526,"about_ca_system_score_gemma":0.0005094805,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00009840947,"about_ca_topic_score_gemma":0.00007614917,"domain_scores_codex":[0.9974939,0.00006169871,0.0003447343,0.001013255,0.0007405921,0.0003458437],"domain_scores_gemma":[0.9977248,0.000103704,0.0002356643,0.001252636,0.0005587306,0.0001245049],"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":[0.0001245451,0.0003528232,0.00002533849,0.04355176,0.0002787516,0.0005375164,0.0127639,0.0009657869,0.001685003,0.02239193,0.06308176,0.8542409],"study_design_scores_gemma":[0.004185528,0.000911969,0.0002932755,0.118287,0.002514813,0.0003048659,0.01075823,0.5791982,0.003024827,0.09533986,0.1779643,0.007217123],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02445733,0.2253618,0.6633616,0.070347,0.00458339,0.005055541,0.0002728379,0.002586387,0.003974079],"genre_scores_gemma":[0.8027064,0.002947231,0.1879178,0.0006064904,0.0001459622,0.000006941825,0.00009913054,0.00004345335,0.005526605],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8470238,"threshold_uncertainty_score":0.9999422,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02300005229061961,"score_gpt":0.2823599422652428,"score_spread":0.2593598899746232,"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."}}