{"id":"W4396832273","doi":"10.1145/3613904.3642669","title":"The ``Colonial Impulse\" of Natural Language Processing: An Audit of Bengali Sentiment Analysis Tools and Their Identity-based Biases","year":2024,"lang":"en","type":"preprint","venue":"","topic":"Hate Speech and Cyberbullying Detection","field":"Computer Science","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"Defense Advanced Research Projects Agency","keywords":"Bengali; Sentiment analysis; Computer science; Sociotechnical system; Artificial intelligence; Colonialism; Identity (music); Natural language processing; Political science; Aesthetics","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":[0.0006994616,0.0002299626,0.0003907622,0.0003316268,0.0001448012,0.0007263977,0.0008218038,0.0001279408,0.00001215853],"category_scores_gemma":[0.00006935476,0.0001425896,0.0002763697,0.0007843418,0.0001155061,0.0002487563,0.000941912,0.0003925537,0.000002109728],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004125707,"about_ca_system_score_gemma":0.0002427083,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0006794443,"about_ca_topic_score_gemma":0.001081798,"domain_scores_codex":[0.9983112,0.0001486928,0.0004196112,0.0005295933,0.0003818134,0.0002090785],"domain_scores_gemma":[0.9985761,0.0001592004,0.0003080414,0.0007191013,0.0001709932,0.00006655648],"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":[0.0001443725,0.0005584673,0.001042482,0.001909725,0.003590832,0.00006468082,0.01412987,0.008078806,0.05968167,0.00557919,0.0003637488,0.9048561],"study_design_scores_gemma":[0.0003107442,0.0003123538,0.00324606,0.0003386406,0.000684058,0.000008129438,0.0008787113,0.7929912,0.1970635,0.003568133,0.000106274,0.0004921889],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9609053,0.003221933,0.03406975,0.0004602278,0.0005990542,0.0003425581,0.00006818915,0.0001703645,0.0001626408],"genre_scores_gemma":[0.9961221,0.00002209779,0.003500876,0.0000336618,0.0000698138,0.00002059886,0.00003312316,0.00001013104,0.0001875995],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.904364,"threshold_uncertainty_score":0.700467,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01767442416322,"score_gpt":0.2905672516684043,"score_spread":0.2728928275051843,"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."}}