{"id":"W3035455711","doi":"10.1007/978-3-030-58323-1_16","title":"Evaluating a Multi-sense Definition Generation Model for Multiple Languages","year":2020,"lang":"en","type":"preprint","venue":"Lecture notes in computer science","topic":"Topic Modeling","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of New Brunswick","funders":"","keywords":"Polysemy; Computer science; Contrast (vision); Word (group theory); Context (archaeology); Natural language processing; Artificial intelligence; Language model; Word-sense disambiguation; Linguistics","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":[0.01228923,0.001401013,0.001605725,0.002308101,0.0009370342,0.004355551,0.002473967,0.00257773,0.004100472],"category_scores_gemma":[0.04195799,0.0007511381,0.002252897,0.001281625,0.0009168045,0.005743318,0.003435523,0.001965747,0.001197138],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002036623,"about_ca_system_score_gemma":0.002423366,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005156623,"about_ca_topic_score_gemma":0.006590006,"domain_scores_codex":[0.989024,0.005361883,0.0007098786,0.001914344,0.002509747,0.000480168],"domain_scores_gemma":[0.9575825,0.03271862,0.001176132,0.003513709,0.003948344,0.001060775],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.004480844,0.001376903,0.01984563,0.001239756,0.001020991,0.0007841214,0.001742858,0.4631714,0.02289831,0.04410855,0.01376839,0.4255623],"study_design_scores_gemma":[0.0000877955,0.0001665262,0.0004979728,0.00002662999,0.0001090912,0.00006613087,0.0001416661,0.9810115,0.003387636,0.01364073,0.0008438558,0.00002039105],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1712828,0.0005561815,0.813751,0.001247898,0.0002421865,0.0004809709,0.001110682,0.008386682,0.002941653],"genre_scores_gemma":[0.6918979,0.0001591457,0.3010531,0.0003332175,0.00009620911,0.0002904776,0.003098374,0.0008911804,0.002180289],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01228923,"threshold_uncertainty_score":0.06499243,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2100460893627557,"score_gpt":0.3725867930988983,"score_spread":0.1625407037361426,"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."}}