{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QOWGFSZFG6ZI37JCNZAUBRMDE2","short_pith_number":"pith:QOWGFSZF","schema_version":"1.0","canonical_sha256":"83ac62cb2537b28dfd226e4140c58326bb3060503e6a6722d3407674b9ea0527","source":{"kind":"arxiv","id":"2605.17710","version":1},"attestation_state":"computed","paper":{"title":"Sometin Beta Pass Notin (SBPN): Improving Multilingual ASR for Nigerian Languages via Knowledge Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Knowledge distillation from monolingual models followed by self-improvement on pseudo labels improves multilingual ASR for Nigerian languages.","cross_cats":["eess.AS"],"primary_cat":"cs.CL","authors_text":"Sewade Ogun","submitted_at":"2026-05-18T00:22:18Z","abstract_excerpt":"Although modern multilingual Automatic Speech Recognition (ASR) systems support several Nigerian languages, their performance consistently lags behind high-resource languages like English and French. Nigerian languages present unique modelling hurdles, including acute data scarcity, inconsistent orthography, tonal diacritics, diverse accents, frequent code-switching, and localized named entities. To address these challenges, we developed a multilingual ASR framework utilizing a two-stage distillation process. First, we employ student-teacher knowledge distillation from existing monolingual mod"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.17710","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-18T00:22:18Z","cross_cats_sorted":["eess.AS"],"title_canon_sha256":"93f4ca82ca769461639ef85d8e0280fe76d22b9b20b0df5e971c6b346f34650d","abstract_canon_sha256":"09722db8d1bdf60f8e279b2cff0ff0db6ef4b09b01f7312f59cb279ad194359b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:54.103946Z","signature_b64":"UWil38vL5kkaQW0/AQCUP4GgCcGADfN+6rOXVELUQAZcIF2hPQ8jpoOfnbVj9LapDWk8PAGOSgTa5MypQqmqBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"83ac62cb2537b28dfd226e4140c58326bb3060503e6a6722d3407674b9ea0527","last_reissued_at":"2026-05-20T00:04:54.103176Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:54.103176Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sometin Beta Pass Notin (SBPN): Improving Multilingual ASR for Nigerian Languages via Knowledge Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Knowledge distillation from monolingual models followed by self-improvement on pseudo labels improves multilingual ASR for Nigerian languages.","cross_cats":["eess.AS"],"primary_cat":"cs.CL","authors_text":"Sewade Ogun","submitted_at":"2026-05-18T00:22:18Z","abstract_excerpt":"Although modern multilingual Automatic Speech Recognition (ASR) systems support several Nigerian languages, their performance consistently lags behind high-resource languages like English and French. Nigerian languages present unique modelling hurdles, including acute data scarcity, inconsistent orthography, tonal diacritics, diverse accents, frequent code-switching, and localized named entities. To address these challenges, we developed a multilingual ASR framework utilizing a two-stage distillation process. First, we employ student-teacher knowledge distillation from existing monolingual mod"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our method significantly bridges the performance gap, achieving on average a relative Word Error Rate (WER) reduction of 29 % over monolingual baselines. Our models also outperform state-of-the-art multilingual models across major benchmarks, including Common Voice and Fleurs.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The iterative self-improvement stage assumes that pseudo-labels generated by the initial distilled model are sufficiently accurate and do not introduce systematic errors that compound over iterations; this premise is invoked when describing the second stage of refinement on pseudo-labelled data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SBPN is a multilingual ASR model for five Nigerian languages that delivers a 29% relative WER reduction over monolingual baselines through teacher-student distillation plus iterative pseudo-label self-improvement.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Knowledge distillation from monolingual models followed by self-improvement on pseudo labels improves multilingual ASR for Nigerian languages.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4bb3121599077e5c3e33b203c5d633fb28a0e4698027702ec04a6fda0fc4ddb4"},"source":{"id":"2605.17710","kind":"arxiv","version":1},"verdict":{"id":"0a6a8fa7-91fd-4cb6-b831-1550e48d0a82","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T22:01:06.027531Z","strongest_claim":"Our method significantly bridges the performance gap, achieving on average a relative Word Error Rate (WER) reduction of 29 % over monolingual baselines. Our models also outperform state-of-the-art multilingual models across major benchmarks, including Common Voice and Fleurs.","one_line_summary":"SBPN is a multilingual ASR model for five Nigerian languages that delivers a 29% relative WER reduction over monolingual baselines through teacher-student distillation plus iterative pseudo-label self-improvement.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The iterative self-improvement stage assumes that pseudo-labels generated by the initial distilled model are sufficiently accurate and do not introduce systematic errors that compound over iterations; this premise is invoked when describing the second stage of refinement on pseudo-labelled data.","pith_extraction_headline":"Knowledge distillation from monolingual models followed by self-improvement on pseudo labels improves multilingual ASR for Nigerian languages."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.17710/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.399565Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T22:10:57.167441Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"shingle_duplication","ran_at":"2026-05-19T21:49:43.475401Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T21:49:43.304918Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.508691Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T21:21:58.548102Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.414497Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"fa437c79dae3535521258ba9aa833e772f41544d169e0f1c79fad2aec525035b"},"references":{"count":31,"sample":[{"doi":"10.21437/interspeech.2023-466","year":2023,"title":"doi: 10.21437/Interspeech.2023-466. Josh Meyer, David Adelani, Edresson Casanova, Alp ¨Oktem, Daniel Whitenack, Julian Weber, Salomon KABONGO KABENAMUALU, Elizabeth Salesky, Iroro Orife, Colin Leong, ","work_id":"6b2f3708-8152-4404-b26e-701631d8e8af","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.18653/v1/2024.emnlp-main.983","year":2024,"title":"In: Proceedings of the 2024 Conference on Em- pirical Methods in Natural Language Processing, EMNLP 2024, Miami, FL, USA, November 12-16","work_id":"c9b74554-3677-4c0a-87dc-5bb9b1a68e21","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1109/taslp.2023.3306709","year":2023,"title":"20 Sometin Beta Pass Notin (SBPN) Appendix A","work_id":"f51842b8-f8dc-41d9-abff-b088c3f3b057","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"beforebefore→bifor bifor","work_id":"a61ec188-80dc-48c1-8543-b5dee0700156","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"chewing gum→ chingum","work_id":"b63824d7-7689-4e82-86dd-fd2465877ba3","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":31,"snapshot_sha256":"468025f93fb317d7e0e0daa205c683259373c808cca120c4c396812dfde92d2d","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"6be5b6b96be6767e2f5cdb9074d82164286f45b323d3d384c4c1a2982fdae5a2"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.17710","created_at":"2026-05-20T00:04:54.103290+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.17710v1","created_at":"2026-05-20T00:04:54.103290+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17710","created_at":"2026-05-20T00:04:54.103290+00:00"},{"alias_kind":"pith_short_12","alias_value":"QOWGFSZFG6ZI","created_at":"2026-05-20T00:04:54.103290+00:00"},{"alias_kind":"pith_short_16","alias_value":"QOWGFSZFG6ZI37JC","created_at":"2026-05-20T00:04:54.103290+00:00"},{"alias_kind":"pith_short_8","alias_value":"QOWGFSZF","created_at":"2026-05-20T00:04:54.103290+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2","json":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2.json","graph_json":"https://pith.science/api/pith-number/QOWGFSZFG6ZI37JCNZAUBRMDE2/graph.json","events_json":"https://pith.science/api/pith-number/QOWGFSZFG6ZI37JCNZAUBRMDE2/events.json","paper":"https://pith.science/paper/QOWGFSZF"},"agent_actions":{"view_html":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2","download_json":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2.json","view_paper":"https://pith.science/paper/QOWGFSZF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.17710&json=true","fetch_graph":"https://pith.science/api/pith-number/QOWGFSZFG6ZI37JCNZAUBRMDE2/graph.json","fetch_events":"https://pith.science/api/pith-number/QOWGFSZFG6ZI37JCNZAUBRMDE2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2/action/storage_attestation","attest_author":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2/action/author_attestation","sign_citation":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2/action/citation_signature","submit_replication":"https://pith.science/pith/QOWGFSZFG6ZI37JCNZAUBRMDE2/action/replication_record"}},"created_at":"2026-05-20T00:04:54.103290+00:00","updated_at":"2026-05-20T00:04:54.103290+00:00"}