Hacker News Reader

Comment by goblin89 | Hacker News Reader

Google should rid itself of the notion that “Traditional Chinese” and “Simplified Chinese” are anything but complete misnomers if you’re talking about languages. They are scripts, and scripts don’t have a 1:1 mapping to languages.

Imagine a world in which Google Translate has one entry called “Latin” for anything using Latin script (Welsh, Irish, you name it) and another entry “Simplified Latin” for English. Just like that, translation “from TC to TC” (e.g., Cantonese to Hakka) is very much a thing!

View on HN

This is not a Latin situation at all, it’s similar to British English vs American English. I just checked and Google Translate doesn’t have separate Englishes for American, British, Australian, New Zealand, Scottish, Indian, etc.

Your comment would be true if American and British English were linguistically classified as distinct, mutually unintelligible languages (beyond amusing accent jokes).

A Mandarin speaker in Taiwan has a different accent and slightly different vocabulary than Mandarin speaker in Beijing, but neither of them would be able to converse with a Yue speaker any more than a Russian speaker would be able to converse with a Polish speaker.

It’s pretty clear Google Translate only outputs Madarin in Simplified and Traditional Chinese. Find a person from either side of the strait, as long as they’re literate and have been taught some version of Madarin, they should understand either version of at least 90% of phrases on the linked page (do a verbatim T2C/C2T first if they have trouble recognizing some characters), only one version may be slightly weird.

Also, you’re very mistaken if you think all versions of English are mutually intelligible.

First, not every Yue speaker speaks Mandarin.

Second, the fact that they also speak Mandarin (because it was a mandatory lesson at school or they were otherwise forced by circumstances to learn it) does not mean there should be no support in Translate for their native language, which incidentally dwarfs many languages that Translate does support in number of speakers.

If you think that’s viable logic, you should try applying it to other languages with predominantly multi-language speakers. Start with Irish (98% of people in Ireland speak English, after all) and move on to Catalan, see where that gets you.

You can support whatever language/dialect you like, doesn’t mean Google Translate has to add it.

You totally can, but we are talking about Google Translate here, how it mislabels scripts as languages and how it doesn’t support a major language (for what is likely political reasons).

The problem is that detecting the different variants to label the training data requires understanding the semantics (detecting French vs English can be done purely lexically), which has historically been a hard problem. With modern language modeling it’s tractable.

I think Google has the resources required to detect different languages in Chinese script (different combinations of characters and vocabulary, even specific Unicode characters, etc.) but there is 1) lack of incentives and 2) opposing political interests. Doesn’t help that there is a bit of a mess when it comes to Chinese characters, like a single Unicode point can be rendered differently depending on the font.

Reply on news.ycombinator.com