{"id":"W2954476116","doi":"10.1109/msr.2019.00046","title":"Can Duplicate Questions on Stack Overflow Benefit the Software Development Community?","year":2019,"lang":"en","type":"article","venue":"","topic":"Software Engineering Research","field":"Computer Science","cited_by":16,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Computer science; Heuristics; Reputation; Information retrieval; Similarity (geometry); Stack (abstract data type); World Wide Web; Data science; Data mining; Artificial intelligence; 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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.01872114,0.0006187512,0.0009245035,0.003826383,0.003018034,0.004256252,0.001466334,0.002844293,0.004216149],"category_scores_gemma":[0.1679918,0.0005477974,0.0006895422,0.002843854,0.002056108,0.01320161,0.005910259,0.001706657,0.001463393],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001485837,"about_ca_system_score_gemma":0.002371155,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003680848,"about_ca_topic_score_gemma":0.005445811,"domain_scores_codex":[0.9786133,0.01058068,0.001418322,0.002256527,0.005841888,0.001289383],"domain_scores_gemma":[0.851664,0.09782714,0.01840033,0.01279445,0.01386479,0.00544925],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.001347666,0.0006829801,0.5084265,0.001648998,0.0002654335,0.003340925,0.09697936,0.002325809,0.01269505,0.01005086,0.03062503,0.3316114],"study_design_scores_gemma":[0.0003504625,0.002520887,0.4421906,0.00131088,0.0005573479,0.01057346,0.1443885,0.0404099,0.02221724,0.05278563,0.2820266,0.0006683456],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9687413,0.001135618,0.01321449,0.004132072,0.00017718,0.0001922963,0.0008154694,0.0009184247,0.01067328],"genre_scores_gemma":[0.9790589,0.0004249744,0.01233473,0.001325405,0.0002550944,0.00009620541,0.00145665,0.0002862625,0.004761782],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9812788,"threshold_uncertainty_score":0.09900808,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02334161124921441,"score_gpt":0.2581692911669891,"score_spread":0.2348276799177747,"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."}}