{"id":"W2805097135","doi":"10.1109/icst.2018.00038","title":"Investigating NLP-Based Approaches for Predicting Manual Test Case Failure","year":2018,"lang":"en","type":"article","venue":"","topic":"Software Engineering Research","field":"Computer Science","cited_by":25,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Calgary","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Test script; Regression testing; Test suite; Test (biology); Test case; Feature selection; Manual testing; Artificial intelligence; Heuristics; Test Management Approach; Feature (linguistics); Software regression; Machine learning; Software; Data mining; Programming language; Software system; Software quality; Software development; Regression analysis; Software construction","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004878044,0.002350251,0.001214724,0.00903872,0.000548564,0.001615997,0.002230056,0.001556069,0.001479382],"category_scores_gemma":[0.02756637,0.000440215,0.001067694,0.00340056,0.000647692,0.002484559,0.0008230371,0.001733577,0.0008083939],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001577846,"about_ca_system_score_gemma":0.001690371,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01769553,"about_ca_topic_score_gemma":0.01592092,"domain_scores_codex":[0.9956449,0.001649473,0.0004143455,0.0008995604,0.001147444,0.0002442597],"domain_scores_gemma":[0.9445125,0.04635607,0.004087144,0.001298206,0.003276067,0.0004699299],"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":[0.000460944,0.001336653,0.1003909,0.001142261,0.0005106943,0.0007408017,0.0007027705,0.3674113,0.009053195,0.003244382,0.004912694,0.5100933],"study_design_scores_gemma":[0.00001984221,0.0001346578,0.00664801,0.00003618033,0.00004701157,0.0001201306,0.0001504264,0.9882578,0.001979805,0.001976354,0.0006002389,0.0000294503],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2001137,0.001875786,0.778304,0.001420042,0.00009215346,0.0006585362,0.002763243,0.01053101,0.004241538],"genre_scores_gemma":[0.7118579,0.0005206754,0.2810764,0.0003361198,0.0001096121,0.0005094159,0.004066797,0.0001998403,0.00132324],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01769553,"threshold_uncertainty_score":0.0351851,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05785945074806147,"score_gpt":0.2848859189331134,"score_spread":0.2270264681850519,"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."}}