{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:X6EHGAKH5YEQ6NHML2KATC22PP","short_pith_number":"pith:X6EHGAKH","schema_version":"1.0","canonical_sha256":"bf88730147ee090f34ec5e94098b5a7be3d60707486ab35baa66545133be0602","source":{"kind":"arxiv","id":"2108.07789","version":2},"attestation_state":"computed","paper":{"title":"Adapting GPT, GPT-2 and BERT Language Models for Speech Recognition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Chao Zhang, Philip C. Woodland, Xianrui Zheng","submitted_at":"2021-07-29T16:53:37Z","abstract_excerpt":"Language models (LMs) pre-trained on massive amounts of text, in particular bidirectional encoder representations from Transformers (BERT), generative pre-training (GPT), and GPT-2, have become a key technology for many natural language processing tasks. In this paper, we present results using fine-tuned GPT, GPT-2, and their combination for automatic speech recognition (ASR). Unlike unidirectional LM GPT and GPT-2, BERT is bidirectional whose direct product of the output probabilities is no longer a valid language prior probability. A conversion method is proposed to compute the correct langu"},"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":"2108.07789","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-07-29T16:53:37Z","cross_cats_sorted":["cs.SD","eess.AS"],"title_canon_sha256":"2afcece8c2ab99d8f3e868abb8b4156f3fdc9c6b1175ded492815cf9ddc06308","abstract_canon_sha256":"fbf64b3c44b56b205fcf0968810dfb6925d565084efa8be4ef60ee2e1eda6abd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:19:20.626077Z","signature_b64":"74p0svqPL6N21isscx4j/YCCiyqizZLwfKgJ7HszpyNYphuOpbIfuGR7yG9rhxvMD8/PFkcpsqg5F0V6dPBTAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bf88730147ee090f34ec5e94098b5a7be3d60707486ab35baa66545133be0602","last_reissued_at":"2026-07-05T03:19:20.625603Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:19:20.625603Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adapting GPT, GPT-2 and BERT Language Models for Speech Recognition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Chao Zhang, Philip C. Woodland, Xianrui Zheng","submitted_at":"2021-07-29T16:53:37Z","abstract_excerpt":"Language models (LMs) pre-trained on massive amounts of text, in particular bidirectional encoder representations from Transformers (BERT), generative pre-training (GPT), and GPT-2, have become a key technology for many natural language processing tasks. In this paper, we present results using fine-tuned GPT, GPT-2, and their combination for automatic speech recognition (ASR). Unlike unidirectional LM GPT and GPT-2, BERT is bidirectional whose direct product of the output probabilities is no longer a valid language prior probability. A conversion method is proposed to compute the correct langu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2108.07789","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/2108.07789/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":"2108.07789","created_at":"2026-07-05T03:19:20.625663+00:00"},{"alias_kind":"arxiv_version","alias_value":"2108.07789v2","created_at":"2026-07-05T03:19:20.625663+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2108.07789","created_at":"2026-07-05T03:19:20.625663+00:00"},{"alias_kind":"pith_short_12","alias_value":"X6EHGAKH5YEQ","created_at":"2026-07-05T03:19:20.625663+00:00"},{"alias_kind":"pith_short_16","alias_value":"X6EHGAKH5YEQ6NHM","created_at":"2026-07-05T03:19:20.625663+00:00"},{"alias_kind":"pith_short_8","alias_value":"X6EHGAKH","created_at":"2026-07-05T03:19:20.625663+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.10058","citing_title":"MTLM: Incorporating Bidirectional Text Information to Enhance Language Model Training in Speech Recognition Systems","ref_index":11,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP","json":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP.json","graph_json":"https://pith.science/api/pith-number/X6EHGAKH5YEQ6NHML2KATC22PP/graph.json","events_json":"https://pith.science/api/pith-number/X6EHGAKH5YEQ6NHML2KATC22PP/events.json","paper":"https://pith.science/paper/X6EHGAKH"},"agent_actions":{"view_html":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP","download_json":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP.json","view_paper":"https://pith.science/paper/X6EHGAKH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2108.07789&json=true","fetch_graph":"https://pith.science/api/pith-number/X6EHGAKH5YEQ6NHML2KATC22PP/graph.json","fetch_events":"https://pith.science/api/pith-number/X6EHGAKH5YEQ6NHML2KATC22PP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP/action/storage_attestation","attest_author":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP/action/author_attestation","sign_citation":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP/action/citation_signature","submit_replication":"https://pith.science/pith/X6EHGAKH5YEQ6NHML2KATC22PP/action/replication_record"}},"created_at":"2026-07-05T03:19:20.625663+00:00","updated_at":"2026-07-05T03:19:20.625663+00:00"}