{"id":"W4412378051","doi":"10.1145/3726302.3730335","title":"Tip of the Tongue Query Elicitation for Simulated Evaluation","year":2025,"lang":"en","type":"article","venue":"","topic":"Speech and dialogue systems","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Microsoft (Canada)","funders":"National Institute of Standards and Technology","keywords":"Computer science; Tongue; Information retrieval; Query optimization; Artificial intelligence; Natural language processing; 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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004484407,0.0000322667,0.00005224313,0.00004269119,0.00003911389,0.00002090924,0.0002304494,0.00002776895,0.000003645415],"category_scores_gemma":[0.0002640023,0.00002085205,0.00003959173,0.0002825724,0.000008503614,0.00008960243,0.0000329164,0.00001455934,0.00000347506],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002212127,"about_ca_system_score_gemma":0.0001127455,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001162834,"about_ca_topic_score_gemma":0.00004932558,"domain_scores_codex":[0.9995192,0.00005496687,0.0001303889,0.00009612025,0.0001384863,0.00006079127],"domain_scores_gemma":[0.9993065,0.0001646761,0.00004619011,0.0002404305,0.0002335928,0.000008585408],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00004529707,0.0001205428,0.004365995,0.0001315023,0.00007472518,1.650967e-7,0.001375343,0.02200804,0.02966485,0.7505155,0.02960537,0.1620927],"study_design_scores_gemma":[0.00057678,0.00002827681,0.01209397,0.00002952327,0.00001068372,2.303742e-7,0.00003992403,0.9048361,0.04510771,0.03583634,0.00139208,0.00004843681],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.09425744,0.00007916216,0.8774682,0.001424048,0.002230199,0.00107987,0.000001929987,0.00006745669,0.02339169],"genre_scores_gemma":[0.9966251,2.460617e-7,0.002060365,0.0002102095,0.00001659702,0.00001426023,0.00000251362,0.000001068846,0.00106963],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9023677,"threshold_uncertainty_score":0.08503218,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01673934936747049,"score_gpt":0.3067126234972936,"score_spread":0.2899732741298231,"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."}}