{"id":"W2786731957","doi":"10.18653/v1/w17-5532","title":"Generating and Evaluating Summaries for Partial Email Threads: Conversational Bayesian Surprise and Silver Standards","year":2017,"lang":"en","type":"article","venue":"","topic":"Advanced Text Analysis Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Computer science; Automatic summarization; Surprise; Thread (computing); Annotation; Redundancy (engineering); Information retrieval; Bayesian probability; Natural language processing; Artificial intelligence; Machine learning; Programming language","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001012201,0.0001204763,0.0001779166,0.00004673914,0.000886096,0.0007466176,0.0002939964,0.00004577616,0.00001595722],"category_scores_gemma":[0.0006452099,0.0001054002,0.00003692072,0.0000331594,0.0001662737,0.001083801,0.0003084065,0.0000533427,2.95091e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003341265,"about_ca_system_score_gemma":0.00009940865,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006399939,"about_ca_topic_score_gemma":0.000228188,"domain_scores_codex":[0.9988732,0.00003411343,0.0001999009,0.00037186,0.0003304732,0.0001903964],"domain_scores_gemma":[0.9988605,0.0001958815,0.0001719461,0.0004255806,0.0002749322,0.00007117276],"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":[0.00007067459,0.00005160378,0.08376414,0.00009268186,0.0001895934,0.000009282322,0.002685405,0.0003003713,0.01359205,0.3790442,0.002566305,0.5176337],"study_design_scores_gemma":[0.0005714644,0.0001043857,0.001659606,0.00001815761,0.00002727066,0.000003970486,0.00005374595,0.9512197,0.01167787,0.03342069,0.001012794,0.0002303873],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04586593,0.0001185832,0.9526403,0.0007014276,0.00004795885,0.0001990991,0.00001600513,0.0001132643,0.0002974353],"genre_scores_gemma":[0.5193498,0.00001841301,0.4803037,0.000101025,0.00006165842,0.0000274362,0.000003813757,0.000005924931,0.0001282775],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9509193,"threshold_uncertainty_score":0.7199651,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03591123173986498,"score_gpt":0.3603061999791383,"score_spread":0.3243949682392733,"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."}}