{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:MTUNM4IVMRTPF2F5CO5NAOSFTZ","short_pith_number":"pith:MTUNM4IV","schema_version":"1.0","canonical_sha256":"64e8d671156466f2e8bd13bad03a459e74fb7b5739aa9409a3d80cfc67904908","source":{"kind":"arxiv","id":"1808.09381","version":2},"attestation_state":"computed","paper":{"title":"Understanding Back-Translation at Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"David Grangier, Michael Auli, Myle Ott, Sergey Edunov","submitted_at":"2018-08-28T16:05:40Z","abstract_excerpt":"An effective method to improve neural machine translation with monolingual data is to augment the parallel training corpus with back-translations of target language sentences. This work broadens the understanding of back-translation and investigates a number of methods to generate synthetic source sentences. We find that in all but resource poor settings back-translations obtained via sampling or noised beam outputs are most effective. Our analysis shows that sampling or noisy synthetic data gives a much stronger training signal than data generated by beam or greedy search. We also compare how"},"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":"1808.09381","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2018-08-28T16:05:40Z","cross_cats_sorted":[],"title_canon_sha256":"8e5d035ade5a01172a90d4f2f7f05accfd0c81654407dd3443c31996a1b851a3","abstract_canon_sha256":"04564badb6c64bddc6570e1d2e4da7ca6a6ac4d5a651fbf3677e84d794c587f1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:04:12.565381Z","signature_b64":"s06eMYtz88DTmiI9xBIvoZt9PkboQRyOJgYJJZxUVee23RxlgP220hwhUALpfLNQTzQyYhy4JkG+SAuqemGIDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"64e8d671156466f2e8bd13bad03a459e74fb7b5739aa9409a3d80cfc67904908","last_reissued_at":"2026-05-18T00:04:12.564709Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:04:12.564709Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding Back-Translation at Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"David Grangier, Michael Auli, Myle Ott, Sergey Edunov","submitted_at":"2018-08-28T16:05:40Z","abstract_excerpt":"An effective method to improve neural machine translation with monolingual data is to augment the parallel training corpus with back-translations of target language sentences. This work broadens the understanding of back-translation and investigates a number of methods to generate synthetic source sentences. We find that in all but resource poor settings back-translations obtained via sampling or noised beam outputs are most effective. Our analysis shows that sampling or noisy synthetic data gives a much stronger training signal than data generated by beam or greedy search. We also compare how"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1808.09381","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":""},"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":"1808.09381","created_at":"2026-05-18T00:04:12.564806+00:00"},{"alias_kind":"arxiv_version","alias_value":"1808.09381v2","created_at":"2026-05-18T00:04:12.564806+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1808.09381","created_at":"2026-05-18T00:04:12.564806+00:00"},{"alias_kind":"pith_short_12","alias_value":"MTUNM4IVMRTP","created_at":"2026-05-18T12:32:40.477152+00:00"},{"alias_kind":"pith_short_16","alias_value":"MTUNM4IVMRTPF2F5","created_at":"2026-05-18T12:32:40.477152+00:00"},{"alias_kind":"pith_short_8","alias_value":"MTUNM4IV","created_at":"2026-05-18T12:32:40.477152+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"1907.05019","citing_title":"Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges","ref_index":4,"is_internal_anchor":true},{"citing_arxiv_id":"2401.02458","citing_title":"Data-Centric Foundation Models in Computational Healthcare: A Survey","ref_index":80,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20912","citing_title":"Enhancing Scientific Discourse: Machine Translation for the Scientific Domain","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"1910.10683","citing_title":"Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2410.13720","citing_title":"Movie Gen: A Cast of Media Foundation Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2005.14165","citing_title":"Language Models are Few-Shot Learners","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18759","citing_title":"Model-Agnostic Meta Learning for Class Imbalance Adaptation","ref_index":46,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ","json":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ.json","graph_json":"https://pith.science/api/pith-number/MTUNM4IVMRTPF2F5CO5NAOSFTZ/graph.json","events_json":"https://pith.science/api/pith-number/MTUNM4IVMRTPF2F5CO5NAOSFTZ/events.json","paper":"https://pith.science/paper/MTUNM4IV"},"agent_actions":{"view_html":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ","download_json":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ.json","view_paper":"https://pith.science/paper/MTUNM4IV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1808.09381&json=true","fetch_graph":"https://pith.science/api/pith-number/MTUNM4IVMRTPF2F5CO5NAOSFTZ/graph.json","fetch_events":"https://pith.science/api/pith-number/MTUNM4IVMRTPF2F5CO5NAOSFTZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ/action/storage_attestation","attest_author":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ/action/author_attestation","sign_citation":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ/action/citation_signature","submit_replication":"https://pith.science/pith/MTUNM4IVMRTPF2F5CO5NAOSFTZ/action/replication_record"}},"created_at":"2026-05-18T00:04:12.564806+00:00","updated_at":"2026-05-18T00:04:12.564806+00:00"}