NamSor public Japanese name translation-training corpus Publisher: NamSor; source: https://github.com/namsor/japanese-names-trans Commit: b04e18e41e9b86089bcf2494da9a718c6595d706 The original repository supplies LGPL-2.1 text in LICENSE. Preserve the included license and applicable notices with covered data. licenses/ONMT_LICENSE.md is the MIT notice for OpenNMT software, not a blanket MIT corpus license. NDR modifications on 2026-11-02T05:24:00.334242+00:00: literal token decoding, frozen normalization/hygiene, deduplication and separate checkpoint corroboration; see MODIFICATIONS.json. Original source files are included byte-for-byte under original/. decoded-approved.tsv is the editable derived name-data source. Records describe use in a public translation-training corpus. Original upstream collection, real-person usage and translation alignment are not independently established for all records. Roles are source labels; independent same-role corroboration is separate. No meaning, gender or language-of-origin claim is made. No API calls, private corpus extraction or model-generated names were used. This is not the proprietary14-billion corpus or a legal guarantee of all upstream rights.