Exactly. We also have to understand that quality is excellent in the case of all bilingual positions and all positions posted in all bilingual communities. We ensure that employers who wish to recruit for positions in the other language receive the best quality.
You say that machine translation is poor. It is true that machine translation is poor if there is no pre-translation, if texts are not repetitive and if there is no revision. A translation machine cannot operate by itself. The technology is not there yet.
We have revised all texts produced by our machine. Now 87% of the texts are good, and that percentage rises to 100% for bilingual communities and bilingual positions. So that leaves the 13% which turns us into a laughing stock, so we will have to continue improving.
The issue is not just money, though money is not a negligible factor. If we dropped machine translation completely, we would have to spend an extra $40 million on job offers posted on a site for two weeks at most. In those circumstances, even the Translation Bureau—and they will tell you this themselves—would use machine translation and revise the text produced. Even the number of positions we post, we simply cannot use translators alone.
Moreover, the texts are identical. When we use pre-translated lists, all employers look at the skill list and take those they need. A translator would lose a great deal of time translating the new list of skills and qualifications. We have reduced the amount of free text. We are preparing pre-translated lists so that the quantity of free text can be reduced even further.