{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:NQJJ62NKB62WKPDXMLWMFTMCKD","short_pith_number":"pith:NQJJ62NK","schema_version":"1.0","canonical_sha256":"6c129f69aa0fb5653c7762ecc2cd8250f15216472ab49f96dbd99ae01c209d44","source":{"kind":"arxiv","id":"1807.07104","version":5},"attestation_state":"computed","paper":{"title":"Hierarchical Multi Task Learning With CTC","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Florian Metze, Ramon Sanabria","submitted_at":"2018-07-18T18:57:37Z","abstract_excerpt":"In Automatic Speech Recognition it is still challenging to learn useful intermediate representations when using high-level (or abstract) target units such as words. For that reason, character or phoneme based systems tend to outperform word-based systems when just few hundreds of hours of training data are being used. In this paper, we first show how hierarchical multi-task training can encourage the formation of useful intermediate representations. We achieve this by performing Connectionist Temporal Classification at different levels of the network with targets of different granularity. Our "},"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":"1807.07104","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2018-07-18T18:57:37Z","cross_cats_sorted":[],"title_canon_sha256":"2d15e05494cc839d95734b63a7fec1ea533cdf27f7d492e5632bd6dbd9cfa7eb","abstract_canon_sha256":"93cca7e3e8ceb3862f3a08eae5f7dadeedb1664201bbc4deafea4a8f60532e69"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:56:29.632190Z","signature_b64":"hoQ+Ure8Tx6z/gpAGQr8+ODdGWhxSvemp27FWUJRLUv9aBYFYoGBLVndTm/PO3Zzphrv7mKwA8aiZjzzMMQ8Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6c129f69aa0fb5653c7762ecc2cd8250f15216472ab49f96dbd99ae01c209d44","last_reissued_at":"2026-05-17T23:56:29.631827Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:56:29.631827Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hierarchical Multi Task Learning With CTC","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Florian Metze, Ramon Sanabria","submitted_at":"2018-07-18T18:57:37Z","abstract_excerpt":"In Automatic Speech Recognition it is still challenging to learn useful intermediate representations when using high-level (or abstract) target units such as words. For that reason, character or phoneme based systems tend to outperform word-based systems when just few hundreds of hours of training data are being used. In this paper, we first show how hierarchical multi-task training can encourage the formation of useful intermediate representations. We achieve this by performing Connectionist Temporal Classification at different levels of the network with targets of different granularity. Our "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1807.07104","kind":"arxiv","version":5},"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":"1807.07104","created_at":"2026-05-17T23:56:29.631881+00:00"},{"alias_kind":"arxiv_version","alias_value":"1807.07104v5","created_at":"2026-05-17T23:56:29.631881+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1807.07104","created_at":"2026-05-17T23:56:29.631881+00:00"},{"alias_kind":"pith_short_12","alias_value":"NQJJ62NKB62W","created_at":"2026-05-18T12:32:40.477152+00:00"},{"alias_kind":"pith_short_16","alias_value":"NQJJ62NKB62WKPDX","created_at":"2026-05-18T12:32:40.477152+00:00"},{"alias_kind":"pith_short_8","alias_value":"NQJJ62NK","created_at":"2026-05-18T12:32:40.477152+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"1907.10726","citing_title":"Cross-Attention End-to-End ASR for Two-Party Conversations","ref_index":39,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD","json":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD.json","graph_json":"https://pith.science/api/pith-number/NQJJ62NKB62WKPDXMLWMFTMCKD/graph.json","events_json":"https://pith.science/api/pith-number/NQJJ62NKB62WKPDXMLWMFTMCKD/events.json","paper":"https://pith.science/paper/NQJJ62NK"},"agent_actions":{"view_html":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD","download_json":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD.json","view_paper":"https://pith.science/paper/NQJJ62NK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1807.07104&json=true","fetch_graph":"https://pith.science/api/pith-number/NQJJ62NKB62WKPDXMLWMFTMCKD/graph.json","fetch_events":"https://pith.science/api/pith-number/NQJJ62NKB62WKPDXMLWMFTMCKD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD/action/storage_attestation","attest_author":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD/action/author_attestation","sign_citation":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD/action/citation_signature","submit_replication":"https://pith.science/pith/NQJJ62NKB62WKPDXMLWMFTMCKD/action/replication_record"}},"created_at":"2026-05-17T23:56:29.631881+00:00","updated_at":"2026-05-17T23:56:29.631881+00:00"}