{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KCXG5LP5BUZSBAOZEMVTN3Z5PU","short_pith_number":"pith:KCXG5LP5","schema_version":"1.0","canonical_sha256":"50ae6eadfd0d332081d9232b36ef3d7d15a57379dc8f8b4303a5fe05bf752bb9","source":{"kind":"arxiv","id":"2601.22792","version":2},"attestation_state":"computed","paper":{"title":"CALM: Joint Contextual Acoustic-Linguistic Modeling for Personalization of Multi-Speaker ASR","license":"http://creativecommons.org/licenses/by/4.0/","headline":"CALM integrates speaker embeddings for target extraction with dynamic vocabulary biasing to halve biased error rates in overlapping multi-speaker ASR.","cross_cats":["cs.CL","cs.SD"],"primary_cat":"eess.AS","authors_text":"Chikara Maeda, Chyi-Jiunn Lin, Muhammad Shakeel, Shinji Watanabe, Yosuke Fukumoto","submitted_at":"2026-01-30T10:12:16Z","abstract_excerpt":"We present CALM, a joint Contextual Acoustic-Linguistic Modeling framework for multi-speaker automatic speech recognition (ASR). In personalized AI scenarios, the joint availability of acoustic and linguistic cues naturally motivates the integration of target-speaker conditioning with contextual biasing in overlapping conversations. CALM implements this integration in an end-to-end framework through speaker embedding-driven target-speaker extraction and dynamic vocabulary-based contextual biasing. We evaluate CALM on simulated English (LibriSpeechMix) and Japanese (Corpus of Spontaneous Japane"},"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":"2601.22792","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"eess.AS","submitted_at":"2026-01-30T10:12:16Z","cross_cats_sorted":["cs.CL","cs.SD"],"title_canon_sha256":"539e452adb9e8cad4111a9bbe956f431d5145aca67b5f57c3d0557c80e80ca06","abstract_canon_sha256":"cd2790a0db990c662e3544ea8ad77ba01427a8f6d561f01a89a1d7d1454cfb20"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:45:05.699300Z","signature_b64":"HrzWkoFy2QLJsP+i4ml4NpKFpb9LR1foekdoTKqp7oHxXII4843xrS1gOSe8Q9VursPBum4GrbYc9hrhRz+kDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"50ae6eadfd0d332081d9232b36ef3d7d15a57379dc8f8b4303a5fe05bf752bb9","last_reissued_at":"2026-05-18T02:45:05.698741Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:45:05.698741Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CALM: Joint Contextual Acoustic-Linguistic Modeling for Personalization of Multi-Speaker ASR","license":"http://creativecommons.org/licenses/by/4.0/","headline":"CALM integrates speaker embeddings for target extraction with dynamic vocabulary biasing to halve biased error rates in overlapping multi-speaker ASR.","cross_cats":["cs.CL","cs.SD"],"primary_cat":"eess.AS","authors_text":"Chikara Maeda, Chyi-Jiunn Lin, Muhammad Shakeel, Shinji Watanabe, Yosuke Fukumoto","submitted_at":"2026-01-30T10:12:16Z","abstract_excerpt":"We present CALM, a joint Contextual Acoustic-Linguistic Modeling framework for multi-speaker automatic speech recognition (ASR). In personalized AI scenarios, the joint availability of acoustic and linguistic cues naturally motivates the integration of target-speaker conditioning with contextual biasing in overlapping conversations. CALM implements this integration in an end-to-end framework through speaker embedding-driven target-speaker extraction and dynamic vocabulary-based contextual biasing. We evaluate CALM on simulated English (LibriSpeechMix) and Japanese (Corpus of Spontaneous Japane"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On two-speaker mixtures, CALM reduces biased word error rate (B-WER) from 12.7 to 4.7 on LibriSpeech2Mix and biased character error rate (B-CER) from 16.6 to 8.4 on CSJMix2 (eval3), demonstrating the effectiveness of joint acoustic-linguistic modeling across languages.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the simulated two-speaker mixtures (LibriSpeechMix, CSJMix) and the IHM-mix condition of AMI sufficiently represent the acoustic and linguistic statistics of real overlapping conversations where speaker turns, noise, and context vary more widely.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CALM jointly models acoustic speaker identity and linguistic context to cut biased error rates by more than half on two-speaker English and Japanese mixtures.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CALM integrates speaker embeddings for target extraction with dynamic vocabulary biasing to halve biased error rates in overlapping multi-speaker ASR.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a2e62ec2c927effb30f1b17bca70f7d9f2104e1d5fdba533710054e6a7c08a08"},"source":{"id":"2601.22792","kind":"arxiv","version":2},"verdict":{"id":"830856c6-8e34-4e42-86b6-e7286d86de38","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T09:47:23.362902Z","strongest_claim":"On two-speaker mixtures, CALM reduces biased word error rate (B-WER) from 12.7 to 4.7 on LibriSpeech2Mix and biased character error rate (B-CER) from 16.6 to 8.4 on CSJMix2 (eval3), demonstrating the effectiveness of joint acoustic-linguistic modeling across languages.","one_line_summary":"CALM jointly models acoustic speaker identity and linguistic context to cut biased error rates by more than half on two-speaker English and Japanese mixtures.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the simulated two-speaker mixtures (LibriSpeechMix, CSJMix) and the IHM-mix condition of AMI sufficiently represent the acoustic and linguistic statistics of real overlapping conversations where speaker turns, noise, and context vary more widely.","pith_extraction_headline":"CALM integrates speaker embeddings for target extraction with dynamic vocabulary biasing to halve biased error rates in overlapping multi-speaker ASR."},"references":{"count":53,"sample":[{"doi":"","year":null,"title":"INTRODUCTION Single-speaker automatic speech recognition (ASR) systems have achieved state-of-the-art (SOTA) performance across many speech- processing tasks [1–3]. However, in multi-speaker settings ","work_id":"dcdd3d30-4ad3-4dd1-9847-c32484f4b15c","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"CALM: Joint Contextual Acoustic-Linguistic Modeling for Personalization of Multi-Speaker ASR","work_id":"76604f8e-13fd-429c-9922-df6583dfac5b","ref_index":2,"cited_arxiv_id":"2601.22792","is_internal_anchor":true},{"doi":"","year":null,"title":"Frame-level target- speaker activity posteriors are computed as: P vad =σ(W vad ˆH(L) +b vad),(11) withP vad ∈[0,1] T enc","work_id":"1b015532-8761-49ec-9892-37edae9d41a3","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2048,"title":"EXPERIMENTS The CALM framework is built on ESPnet [45], pairing a Conformer encoder with a Transformer decoder. The Conformer has 12 lay- ers with 4 heads and 1024 linear units (kernel size 31) and ap","work_id":"1d55f36d-0f28-4a57-b4eb-7773cdc825ce","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"However, unlike in simulated conditions, overall WER in- creases from 37.4 to 39.1 absolute points. Our error analysis indi- cates that this degradation is primarily driven by an increase in inser- ti","work_id":"020dd479-0292-4a34-b2b0-c20ff1f0891f","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":53,"snapshot_sha256":"44546e885cd008a81f565798e10d28d5b34db6c030c1b4dfc303290113d923a6","internal_anchors":1},"formal_canon":{"evidence_count":1,"snapshot_sha256":"372247d71b14197b1b40475030f8ccb99c8b2aeb05830ab6a298b23308cf8939"},"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":"2601.22792","created_at":"2026-05-18T02:45:05.698853+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.22792v2","created_at":"2026-05-18T02:45:05.698853+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.22792","created_at":"2026-05-18T02:45:05.698853+00:00"},{"alias_kind":"pith_short_12","alias_value":"KCXG5LP5BUZS","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"KCXG5LP5BUZSBAOZ","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"KCXG5LP5","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2601.22792","citing_title":"CALM: Joint Contextual Acoustic-Linguistic Modeling for Personalization of Multi-Speaker ASR","ref_index":2,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":1,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU","json":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU.json","graph_json":"https://pith.science/api/pith-number/KCXG5LP5BUZSBAOZEMVTN3Z5PU/graph.json","events_json":"https://pith.science/api/pith-number/KCXG5LP5BUZSBAOZEMVTN3Z5PU/events.json","paper":"https://pith.science/paper/KCXG5LP5"},"agent_actions":{"view_html":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU","download_json":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU.json","view_paper":"https://pith.science/paper/KCXG5LP5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.22792&json=true","fetch_graph":"https://pith.science/api/pith-number/KCXG5LP5BUZSBAOZEMVTN3Z5PU/graph.json","fetch_events":"https://pith.science/api/pith-number/KCXG5LP5BUZSBAOZEMVTN3Z5PU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU/action/storage_attestation","attest_author":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU/action/author_attestation","sign_citation":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU/action/citation_signature","submit_replication":"https://pith.science/pith/KCXG5LP5BUZSBAOZEMVTN3Z5PU/action/replication_record"}},"created_at":"2026-05-18T02:45:05.698853+00:00","updated_at":"2026-05-18T02:45:05.698853+00:00"}