80% accuracy is really bad for this. People are really offended if you get their gender wrong and they notice, especially if, like most people, they haven’t been worn down by that being an everyday occurrence.
It’s an interesting description of a machine learning technique, I just know that there are companies using things like it, and am deeply saddened by that.
Agreed. If you think you need gender information, you probably don’t.
Well, you have to pick a pronoun at some point. You can use ‘they’ but there’s also people who get wildly upset at that for no good reason. Pick your poison.
Why? The only reason I’ve run into for software to generate text referring to users in the third-person is when it’s some sort of social networking or communication tool. And when it’s that, users should be specifying their pronouns (as Facebook and Google+ both allow; Twitter doesn’t, but avoids pronouns entirely).
There are languages where second-person pronouns have to be gendered, and that must indeed be awkward wrt addressing the user. I tend to favor messaging that avoids trying to greet the user familiarly, which would avoid this issue… but I am told it has great appeal from a marketing standpoint, and I’m not equipped to argue with that.
Oh, I wasn’t thinking of just software. Missed the context.
The thing is, sometimes ‘they’ is the correct pronoun, and yet when someone says “you have to pick a pronoun” they rarely mean to include that as one of the options.
Better to ask for pronouns if you need pronouns, and not gender information.
I never used to use “they” because I was taught to use “he or she” but the singular “they” is becoming accepted by style guides and the like, so I feel less weird using it.
If it makes you feel any better, singular they was common usage until 20th-century prescriptive grammarians decided it shouldn’t be. Most people still use it, but only in relaxed speech - if you point out they’ve been doing it, it’s common for people to switch to “he or she”. What we’re told at age ten has this way of sticking with us…
this is one place where us southerners do it right - when in doubt, just say y'all. It works for second person plural, singular, and I’ve even heard enterprising folks stretch it into the third person.
Interesting - i would also like to see how an approach based on maximum common fragments would work. You build a dictionary of fragments of names, and try to score based on the similarity or occurrence of those.
E.g : AND® - is a fragment that comes from the Greek andros (man, manly ) and should have a higher M occurrence (Alexander etc) . However you also have one of the most common counter examples in andrea.
I’d like to see how many of those deviations occur..
Rambling apart, Then I’m sadly a male owner of a female name (here in US) so there’s nothing you can do to match me, unless you take also nationalities into account.
Yeah, im not sure you can solve those cases. Age 30, country America, name Jordan, which gender? You’re just never going to win with these sort of ambiguous names. Well, not more than almost 50%.
I think it would be nice to use these fancy algorithms to find me and my pals some good ambiguous nb names tbh
They don’t seem to be making any effort to detect non-binary genders, or even mention that there would be some inevitable error because their model is incomplete.
I suspect they don’t even know their model is incomplete, because they don’t bother to do any research before jumping ahead into such a fraught area.