Synthetic Datasets
Identity
Description
100 hand-curated false friends (20 each in Norwegian, German, Dutch, Spanish and Portuguese), in words that look like English but don't mean what you'd think. Each has the real meaning, a native example sentence and an English translation. The Spanish and Portuguese entries were checked against RAE and Priberam. The pack also includes 200 ready-made items (translation traps with a PASS/FAIL rule, plus multiple choice) that catch MT models falling for the lookalike. JSONL and CSV.
Tags
dataset, false-friends, machine-translation, mt-evaluation, glossary, norwegian, german, dutch, spanish, portuguese, csv, jsonl
Technical
Schema format
Json Schema
Industry
technology
File size
25 KB