or
Pick a sample and press Convert.
What has been converted here
The same 22 documents under each translator, each round of the recogniser and each reader tried in its place, grouped by document; the best reading of each is marked. Pick one model to rank its runs by score. Open a row to see its stages and its score again, without paying for the conversion twice. The means line above the list has the averages.
Round by round
How the models compare
The 22-document benchmark: each translator over each round of the recogniser, and each reader tried in the recogniser's place, scored by one judge reading the delivered text against the page three times.
loading…