2 ms·
OK, but this still has the same problem as BLEU- it relies on comparisons to human scores, which are entirely subjective. I'm not saying they're not the best we
by YeGoblynQueenne 9y ago
OK, but this still has the same problem as BLEU- it relies on comparisons to human scores, which are entirely subjective. I'm not saying they're not the best we got, but it's a big problem for machine translation that the only way to evaluate results is, essentially, comparing it to eyballing.