{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:MT7A5PFRB4QRRA7PNJD666ZKOU","short_pith_number":"pith:MT7A5PFR","schema_version":"1.0","canonical_sha256":"64fe0ebcb10f211883ef6a47ef7b2a7511de1a3bc050cb7f2f9e0cf5f7adba2d","source":{"kind":"arxiv","id":"2106.02302","version":1},"attestation_state":"computed","paper":{"title":"Minimum Word Error Rate Training with Language Model Fusion for End-to-End Speech Recognition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG","cs.SD"],"primary_cat":"eess.AS","authors_text":"Eric Sun, Guoli Ye, Jinyu Li, Liang Lu, Naoyuki Kanda, Xie Chen, Yifan Gong, Yu Wu, Zhong Meng","submitted_at":"2021-06-04T07:24:49Z","abstract_excerpt":"Integrating external language models (LMs) into end-to-end (E2E) models remains a challenging task for domain-adaptive speech recognition. Recently, internal language model estimation (ILME)-based LM fusion has shown significant word error rate (WER) reduction from Shallow Fusion by subtracting a weighted internal LM score from an interpolation of E2E model and external LM scores during beam search. However, on different test sets, the optimal LM interpolation weights vary over a wide range and have to be tuned extensively on well-matched validation sets. In this work, we perform LM fusion in "},"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":"2106.02302","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"eess.AS","submitted_at":"2021-06-04T07:24:49Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG","cs.SD"],"title_canon_sha256":"630f278f493f8ce5b0a8ff2eeaa111767ffb9b288fd68a8a2862cc25e83de50d","abstract_canon_sha256":"ecf9328c610e404c8ee979e93159a92fbb057bc7be730266e0bd602b0c48b78d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:46:17.921917Z","signature_b64":"BkeC7v6Y1bVkVGSNJQjosY1TrxZ2F2giU72DAdOm1aXYYrwpNQb2PefWaOz+dP8HIzudZNXJNX4XCPROvDLCCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"64fe0ebcb10f211883ef6a47ef7b2a7511de1a3bc050cb7f2f9e0cf5f7adba2d","last_reissued_at":"2026-07-05T02:46:17.921438Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:46:17.921438Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Minimum Word Error Rate Training with Language Model Fusion for End-to-End Speech Recognition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG","cs.SD"],"primary_cat":"eess.AS","authors_text":"Eric Sun, Guoli Ye, Jinyu Li, Liang Lu, Naoyuki Kanda, Xie Chen, Yifan Gong, Yu Wu, Zhong Meng","submitted_at":"2021-06-04T07:24:49Z","abstract_excerpt":"Integrating external language models (LMs) into end-to-end (E2E) models remains a challenging task for domain-adaptive speech recognition. Recently, internal language model estimation (ILME)-based LM fusion has shown significant word error rate (WER) reduction from Shallow Fusion by subtracting a weighted internal LM score from an interpolation of E2E model and external LM scores during beam search. However, on different test sets, the optimal LM interpolation weights vary over a wide range and have to be tuned extensively on well-matched validation sets. In this work, we perform LM fusion in "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.02302","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/2106.02302/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":"2106.02302","created_at":"2026-07-05T02:46:17.921491+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.02302v1","created_at":"2026-07-05T02:46:17.921491+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.02302","created_at":"2026-07-05T02:46:17.921491+00:00"},{"alias_kind":"pith_short_12","alias_value":"MT7A5PFRB4QR","created_at":"2026-07-05T02:46:17.921491+00:00"},{"alias_kind":"pith_short_16","alias_value":"MT7A5PFRB4QRRA7P","created_at":"2026-07-05T02:46:17.921491+00:00"},{"alias_kind":"pith_short_8","alias_value":"MT7A5PFR","created_at":"2026-07-05T02:46:17.921491+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU","json":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU.json","graph_json":"https://pith.science/api/pith-number/MT7A5PFRB4QRRA7PNJD666ZKOU/graph.json","events_json":"https://pith.science/api/pith-number/MT7A5PFRB4QRRA7PNJD666ZKOU/events.json","paper":"https://pith.science/paper/MT7A5PFR"},"agent_actions":{"view_html":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU","download_json":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU.json","view_paper":"https://pith.science/paper/MT7A5PFR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.02302&json=true","fetch_graph":"https://pith.science/api/pith-number/MT7A5PFRB4QRRA7PNJD666ZKOU/graph.json","fetch_events":"https://pith.science/api/pith-number/MT7A5PFRB4QRRA7PNJD666ZKOU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU/action/storage_attestation","attest_author":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU/action/author_attestation","sign_citation":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU/action/citation_signature","submit_replication":"https://pith.science/pith/MT7A5PFRB4QRRA7PNJD666ZKOU/action/replication_record"}},"created_at":"2026-07-05T02:46:17.921491+00:00","updated_at":"2026-07-05T02:46:17.921491+00:00"}