{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3Z3JZTZJU2WA2JHDLWUNJ3ICAH","short_pith_number":"pith:3Z3JZTZJ","schema_version":"1.0","canonical_sha256":"de769ccf29a6ac0d24e35da8d4ed0201c7d763af083a1b19ed6202da2899760b","source":{"kind":"arxiv","id":"2502.10058","version":2},"attestation_state":"computed","paper":{"title":"MTLM: Incorporating Bidirectional Text Information to Enhance Language Model Training in Speech Recognition Systems","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["eess.AS"],"primary_cat":"cs.CL","authors_text":"Changsong Dai, Huizhi Liang, Pengju Ren, Qingliang Meng, Tian Li","submitted_at":"2025-02-14T10:21:10Z","abstract_excerpt":"Automatic speech recognition (ASR) systems normally consist of an acoustic model (AM) and a language model (LM). The acoustic model estimates the probability distribution of text given the input speech, while the language model calibrates this distribution toward a specific knowledge domain to produce the final transcription. Traditional ASR-specific LMs are typically trained in a unidirectional (left-to-right) manner to align with autoregressive decoding. However, this restricts the model from leveraging the right-side context during training, limiting its representational capacity. In this w"},"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":"2502.10058","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-14T10:21:10Z","cross_cats_sorted":["eess.AS"],"title_canon_sha256":"6ab3bd5d04c38cd271df72357f9bafb5a5ce039ed54a4a58193637bd692eda85","abstract_canon_sha256":"481efa27d0caebf3e89606ab5632117393406458e6ff58d986cf88019ac693ae"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:21:42.264210Z","signature_b64":"lQNLB6rcy3r5FRJraU7Vthocq+viVpxAw80lTxD9XDIc9eVWvFz/kx3GU5ueJWioOuS/6VxMYqRemDPrjUaTCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"de769ccf29a6ac0d24e35da8d4ed0201c7d763af083a1b19ed6202da2899760b","last_reissued_at":"2026-07-05T11:21:42.263716Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:21:42.263716Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MTLM: Incorporating Bidirectional Text Information to Enhance Language Model Training in Speech Recognition Systems","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["eess.AS"],"primary_cat":"cs.CL","authors_text":"Changsong Dai, Huizhi Liang, Pengju Ren, Qingliang Meng, Tian Li","submitted_at":"2025-02-14T10:21:10Z","abstract_excerpt":"Automatic speech recognition (ASR) systems normally consist of an acoustic model (AM) and a language model (LM). The acoustic model estimates the probability distribution of text given the input speech, while the language model calibrates this distribution toward a specific knowledge domain to produce the final transcription. Traditional ASR-specific LMs are typically trained in a unidirectional (left-to-right) manner to align with autoregressive decoding. However, this restricts the model from leveraging the right-side context during training, limiting its representational capacity. In this w"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.10058","kind":"arxiv","version":2},"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/2502.10058/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":"2502.10058","created_at":"2026-07-05T11:21:42.263777+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.10058v2","created_at":"2026-07-05T11:21:42.263777+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.10058","created_at":"2026-07-05T11:21:42.263777+00:00"},{"alias_kind":"pith_short_12","alias_value":"3Z3JZTZJU2WA","created_at":"2026-07-05T11:21:42.263777+00:00"},{"alias_kind":"pith_short_16","alias_value":"3Z3JZTZJU2WA2JHD","created_at":"2026-07-05T11:21:42.263777+00:00"},{"alias_kind":"pith_short_8","alias_value":"3Z3JZTZJ","created_at":"2026-07-05T11:21:42.263777+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.08477","citing_title":"ILT-Iterative LoRA Training through Focus-Feedback-Fix for Multilingual Speech Recognition","ref_index":35,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH","json":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH.json","graph_json":"https://pith.science/api/pith-number/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/graph.json","events_json":"https://pith.science/api/pith-number/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/events.json","paper":"https://pith.science/paper/3Z3JZTZJ"},"agent_actions":{"view_html":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH","download_json":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH.json","view_paper":"https://pith.science/paper/3Z3JZTZJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.10058&json=true","fetch_graph":"https://pith.science/api/pith-number/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/graph.json","fetch_events":"https://pith.science/api/pith-number/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/action/storage_attestation","attest_author":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/action/author_attestation","sign_citation":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/action/citation_signature","submit_replication":"https://pith.science/pith/3Z3JZTZJU2WA2JHDLWUNJ3ICAH/action/replication_record"}},"created_at":"2026-07-05T11:21:42.263777+00:00","updated_at":"2026-07-05T11:21:42.263777+00:00"}