{"id":"W3154790171","doi":"10.18653/v1/2021.eacl-main.243","title":"Measuring and Improving Faithfulness of Attention in Neural Machine Translation","year":2021,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":14,"is_retracted":false,"has_abstract":true,"ca_institutions":"Simon Fraser University","funders":"Natural Sciences and Engineering Research Council of Canada; Ministère de la Défense Nationale","keywords":"Machine translation; Computer science; Regularization (linguistics); Measure (data warehouse); Artificial intelligence; Translation (biology); Divergence (linguistics); Differentiable function; Deep neural networks; Quality (philosophy); Machine learning; Artificial neural network; Natural language processing; Mathematics; Data mining; Linguistics","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":[0.005696949,0.001098839,0.0009681899,0.001125034,0.0006696889,0.001789143,0.001276009,0.001708116,0.001222173],"category_scores_gemma":[0.03782788,0.0006613715,0.0007954987,0.000816849,0.001856015,0.003535713,0.002562092,0.002654371,0.0002669743],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001437822,"about_ca_system_score_gemma":0.0006851364,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002511387,"about_ca_topic_score_gemma":0.002612459,"domain_scores_codex":[0.9975572,0.001313001,0.0001672845,0.0004966253,0.0003285523,0.0001373299],"domain_scores_gemma":[0.980824,0.01323073,0.001224412,0.00311953,0.001205342,0.0003959349],"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":[0.001091932,0.0003004405,0.01198461,0.0002725525,0.0002970301,0.0002282394,0.001014001,0.7443017,0.027034,0.01892163,0.001407976,0.1931458],"study_design_scores_gemma":[0.00002504173,0.0001780852,0.001894808,0.00001914211,0.00003274047,0.00005339369,0.00004960529,0.9679661,0.007391741,0.02208664,0.0002781881,0.00002444756],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4020247,0.0007221191,0.5925781,0.0007237965,0.00006497077,0.00007323889,0.0001136447,0.001800935,0.001898516],"genre_scores_gemma":[0.9573733,0.00009629394,0.04140254,0.0001401916,0.00003113683,0.00004767291,0.0001418766,0.0001813361,0.0005855748],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005696949,"threshold_uncertainty_score":0.03012872,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04442249320317228,"score_gpt":0.2302141650704904,"score_spread":0.1857916718673182,"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."}}