This package contains a ready-made Quantum Machine Translation™ (QMT™) engine to import and immediately use in Slate Desktop™, Slate Desktop Pro™ and Slate Connect™.
| Engine Name | en-ro-ep7_20191011q |
| Subject/specialty | Generic, parliament debates |
| Source language | English (en) |
| Target language | Romanian (ro) |
| BLEU score | 59.95037 |
| Quality quotient (QQ) | 15.9% |
| Edit distance per line (Levenshtein) | 34.2 |
| Evaluation segments (count) | 2,345 |
| BLEU 1.0 segments (count) | 373 |
| Evaluation set (free download) | Download Free |
Training Resources
This package includes the engine’s training corpus. We used the QMT strategy to create the smallest possible training corpus that maximizes machine translation post-editing productivity.
| Parallel Training Corpus | ep7_20191011q |
| Source/target pairs (count) | 103,006 |
| Lexicon, source language (words) | 30,142 |
| Words, source language (total) | 1,985,082 |
| Lexicon, target language (words) | 51,025 |
| Words, target language (total) | 1,976,272 |
| Monolingual Training Corpus | ep7_20191011q |
| Target segments (count) | 112,735 |
| Lexicon, target language (words) | 54,994 |
| Words, target language (total) | 2,365,217 |
| Original EuroParl v7.0 total segment pairs | 399,375 |