3 ms·
> "(name) (Roman numeral)" is a much more commong occurence than "(name) (Single letter spelled out)" > So to answer your question, this one. This universe. B
by wtmt 5y ago
> "(name) (Roman numeral)" is a much more commong occurence than "(name) (Single letter spelled out)"
> So to answer your question, this one. This universe.
But this is true in this universe only in some countries and places. There are many countries and places where an initial would follow the name and it is the legal name.
This is precisely the problem with the data/algorithm as well as your argument. They’re not well formed for different contexts and need to have more context sensitivity.
- refenestrator 5y agoCultural sensitivity, you mean. If it mispronounced a random non-icon's name everyone would chuckle and forget about it. Pattern matchers don't understand culture, so this is gonna be a hard-coded workaround or intentional distortion of dataset to try and achieve the hard-coded workaround.
- scrollaway 5y agoThe comment i was replying to straight up claimed this was done with intent. I agree with you, the algorithm needs to have more context sensitivity as you say. But nobody at google is getting their rocks off making the AI mispronounce things on purpose. Especially since, as anyone who used Google Maps in multilingual areas will attest, the algorithm is awful at dealing with context in general. It mispronounces tons of things. In Belgium it will switch between dutch and french names for streets randomly. Google Maps canonicalizes my address in Dutch, so whenever i input it on a website i have to deal with it "helpfully" "correcting" it to dutch. Which makes things difficult when I have to deal with French speakers... Google Maps has a lot of problems that are causing real harm to non-us cultures, which would likely get fixed with diversity. But i would never claim these things are done on purpose.