{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PG2MUXECQ7VLMBTIYWHWR7PSFK","short_pith_number":"pith:PG2MUXEC","schema_version":"1.0","canonical_sha256":"79b4ca5c8287eab60668c58f68fdf22a867d1eb4fbe271da2502222a5e92cf51","source":{"kind":"arxiv","id":"2407.16664","version":1},"attestation_state":"computed","paper":{"title":"Towards scalable efficient on-device ASR with transfer learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["eess.AS"],"primary_cat":"cs.CL","authors_text":"Arthur Guo, Debjyoti Paul, Jay Mahadeokar, Jinxi Guo, Ke Li, Laxmi Pandey, Xuedong Zhang","submitted_at":"2024-07-23T17:29:02Z","abstract_excerpt":"Multilingual pretraining for transfer learning significantly boosts the robustness of low-resource monolingual ASR models. This study systematically investigates three main aspects: (a) the impact of transfer learning on model performance during initial training or fine-tuning, (b) the influence of transfer learning across dataset domains and languages, and (c) the effect on rare-word recognition compared to non-rare words. Our finding suggests that RNNT-loss pretraining, followed by monolingual fine-tuning with Minimum Word Error Rate (MinWER) loss, consistently reduces Word Error Rates (WER)"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2407.16664","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-23T17:29:02Z","cross_cats_sorted":["eess.AS"],"title_canon_sha256":"2d9742b04e18002294b9a1a5ded906abda672a45a3caa898686aa5d33101ff4f","abstract_canon_sha256":"6e3cc10310189dfb0d169c2aa652786204bb88f2de95eb8a1049556321bc51f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:47:36.686174Z","signature_b64":"G4gmp5Ab2w3Tpv0SmkYp6xWaeXAEKJFoLw3QvMXlRTSoDzvTyxqtVrMCz9ljW9gHaAhYBNMq9hl3Xkoh6vcIBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"79b4ca5c8287eab60668c58f68fdf22a867d1eb4fbe271da2502222a5e92cf51","last_reissued_at":"2026-07-05T08:47:36.685723Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:47:36.685723Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards scalable efficient on-device ASR with transfer learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["eess.AS"],"primary_cat":"cs.CL","authors_text":"Arthur Guo, Debjyoti Paul, Jay Mahadeokar, Jinxi Guo, Ke Li, Laxmi Pandey, Xuedong Zhang","submitted_at":"2024-07-23T17:29:02Z","abstract_excerpt":"Multilingual pretraining for transfer learning significantly boosts the robustness of low-resource monolingual ASR models. This study systematically investigates three main aspects: (a) the impact of transfer learning on model performance during initial training or fine-tuning, (b) the influence of transfer learning across dataset domains and languages, and (c) the effect on rare-word recognition compared to non-rare words. Our finding suggests that RNNT-loss pretraining, followed by monolingual fine-tuning with Minimum Word Error Rate (MinWER) loss, consistently reduces Word Error Rates (WER)"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.16664","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2407.16664/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"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":"2407.16664","created_at":"2026-07-05T08:47:36.685788+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.16664v1","created_at":"2026-07-05T08:47:36.685788+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.16664","created_at":"2026-07-05T08:47:36.685788+00:00"},{"alias_kind":"pith_short_12","alias_value":"PG2MUXECQ7VL","created_at":"2026-07-05T08:47:36.685788+00:00"},{"alias_kind":"pith_short_16","alias_value":"PG2MUXECQ7VLMBTI","created_at":"2026-07-05T08:47:36.685788+00:00"},{"alias_kind":"pith_short_8","alias_value":"PG2MUXEC","created_at":"2026-07-05T08:47:36.685788+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24169","citing_title":"Data Scale, Not Latency, Shapes Cross-Lingual Encoder Transfer in Streaming ASR","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK","json":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK.json","graph_json":"https://pith.science/api/pith-number/PG2MUXECQ7VLMBTIYWHWR7PSFK/graph.json","events_json":"https://pith.science/api/pith-number/PG2MUXECQ7VLMBTIYWHWR7PSFK/events.json","paper":"https://pith.science/paper/PG2MUXEC"},"agent_actions":{"view_html":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK","download_json":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK.json","view_paper":"https://pith.science/paper/PG2MUXEC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.16664&json=true","fetch_graph":"https://pith.science/api/pith-number/PG2MUXECQ7VLMBTIYWHWR7PSFK/graph.json","fetch_events":"https://pith.science/api/pith-number/PG2MUXECQ7VLMBTIYWHWR7PSFK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK/action/storage_attestation","attest_author":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK/action/author_attestation","sign_citation":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK/action/citation_signature","submit_replication":"https://pith.science/pith/PG2MUXECQ7VLMBTIYWHWR7PSFK/action/replication_record"}},"created_at":"2026-07-05T08:47:36.685788+00:00","updated_at":"2026-07-05T08:47:36.685788+00:00"}