At a quick glance, the translated portion looks good, other than one fairly major error (lacets translates as “loops”, not “braids” - I call it fairly major because “braids” have a different topological meaning). I agree with the objections about the introduction generally being easier to translate. I’m trying to find a portion of the paper that would be fairly tricky, but that’s somewhat difficult in this case – one reason this paper is considered a classic in the field is that it’s well-written.
Perhaps Itself can come up with something - perhaps involving faisceaux?
Thanks for the correction - I would call that a major error as well. Given that I’m not a translator with experience in working with mathematics, and that this is a ‘proof of concept’ exercise rather than coursework or an accreditation exam, how tricky do you feel it needs to be? The fact that it’s well-written isn’t an advantage in this case, as it’s the unfamiliar terminology that’s the bigger obstacle here.
Frankly, in order for this test to be meaningful at all, it needs to be carried out on the most mathematically and conceptually “tricky” passages that can be found. No-one doubts that a competent translator can translate the straightforward stuff, or, indeed, that they will be able to look up unfamiliar terms. The issue is whether a translator who is not also an expert mathematician will make errors when it comes to more conceptually difficult points where the intended meaning can only be disambiguated (even when reading in one’s native language) by someone who actually understands the mathematical argument that is being made.