PiotrSty's picture
setimes_tr v1: 201154 TR-EN pairs, 11.0M tokens
401912c verified
Raw History Blame Contribute Delete
2.32 kB
{
"scope": "Turkish-English sentence pairs from the OPUS SETIMES v2 Moses package (SETIMES.en-tr, release 2018-03-04)",
"rejection_counts": {
"low_letter_ratio": 1906,
"near_duplicate": 10,
"non_turkish_text": 2226,
"normalized_duplicate": 1791,
"too_little_text": 591
},
"pii_pattern_matches": {
"email": 4,
"labelled_phone": 0,
"ipv4": 0,
"ipv6": 0,
"national_identifier": 0,
"iban": 0
},
"residual_pattern_document_counts": {
"email": 0,
"labelled_phone": 0,
"ipv4": 0,
"ipv6": 0,
"national_identifier": 0,
"iban": 0
},
"pii_limitations": "Pattern redaction only (labelled patterns: e-mail, labelled phone, labelled IP, labelled national identifier, IBAN candidates), applied to both language sides. Bare dotted quads are content. Not anonymization - news text names real persons as published.",
"license_gate": "CC-BY-SA as stated by OPUS (opus.nlpl.eu/datasets/SETIMES: 'License CC-BY-SA license'); the package LICENSE defers to the original setimes.com sources ('The data set comes with the same license as the original sources'). Accepted with provenance - release: None pending review (the ted-polish NOTICE precedent).",
"language_gate": "langid == tr on the first 1000 characters of the Turkish side",
"minimum_final_text_characters": 20,
"exact_dedup": "128-bit truncated sha256 of final Turkish text",
"near_dedup": "Jaccard >= 0.9 on capped 25-shingle sets (>= 20 shared shingles) - inert for sentence-length records, which rely on exact dedup",
"selection": "all aligned pairs of the Moses export (record = sentence pair; the .ids file is per-pair alignment provenance, not document structure)",
"date_extraction": "created parsed best-effort from SETimes article datelines ('-- dd/mm/yy') present in some sentences",
"layout_filter": "drop_layout_lines (1 line instances removed)",
"boilerplate": "lines binned to 67108864 64-bit hash bins with document frequency > 2029 dropped (0 line instances removed; max bin df 149)",
"limitations": "Sentence-pair record model (news sentences, 2000-2013 SETimes coverage); MT-unfriendly alignment artifacts possible; byline datelines cover only a fraction of records; near-dedup does not fire on sentence-length rows."
}