Live data from Hacker News

Identify Any Language Written in the Roman Alphabet at a Glance

theweek.com

61–70 of 95 posts

Re: Identify Any Language Written in the Roman Alphabet at a Glance

#61
post #22
post #2

This is all very helpful, but how do you spot English?

No diacritics, and generous use of apostrophes.

That guideline would likely classify Xhosa as English. See https://xh.wikipedia.org/wiki/Iphepha_Elingundoqo , which has no diacritics and three apostrophes.

Re: Identify Any Language Written in the Roman Alphabet at a Glance

#63
post #30
post #5

Additional bonus: Korean uses lots of basic angles, squares, and lines and many Hangul have "three parts" eg. 한국어의 It's a beautifully simple writing system. You can go quite a long comment chain in Japanese without seeing の. I always tell people "look for lots of simple characters that can be written with only 2 or 3 lines mixed in between a bunch of really complex characters". 来週は学校に行きません For those who don't know Ja…

I would say the give away on Korean is that it has circles or ovals as part of the character.

It doesn't necessarily have any though.

Re: Identify Any Language Written in the Roman Alphabet at a Glance

#64
post #56
post #37

Earlier quoted context omitted.

SwiftKey (Android keyboard) does an amazing job of this. I routinely switch between French and English, and usually by the end of the first word in the other language my autocorrect and suggestions are both in the right language.

AFAIK it uses Markov chains, so I think it just lumps all dictionaries together, and as soon as you write a couple of words, the probabilities for the following words will be in the proper language. It doesn't even need specific rules per language, it's all automatic.

That is my guess as well but, whatever it is, it works marvelously. Many times, I write a significant portion of my message just through suggestions.

Re: Identify Any Language Written in the Roman Alphabet at a Glance

#65
post #56
post #37

Earlier quoted context omitted.

SwiftKey (Android keyboard) does an amazing job of this. I routinely switch between French and English, and usually by the end of the first word in the other language my autocorrect and suggestions are both in the right language.

AFAIK it uses Markov chains, so I think it just lumps all dictionaries together, and as soon as you write a couple of words, the probabilities for the following words will be in the proper language. It doesn't even need specific rules per language, it's all automatic.

There's now SwiftKey Neural, that probably uses Neuronal Nets instead of markov chains.

Re: Identify Any Language Written in the Roman Alphabet at a Glance

#66
post #64
post #56

Earlier quoted context omitted.

AFAIK it uses Markov chains, so I think it just lumps all dictionaries together, and as soon as you write a couple of words, the probabilities for the following words will be in the proper language. It doesn't even need specific rules per language, it's all automatic.

That is my guess as well but, whatever it is, it works marvelously. Many times, I write a significant portion of my message just through suggestions.

That's a funny game to play with your friends (if they also have personalized suggestions). Just see what you can write using only your suggestions.

Re: Identify Any Language Written in the Roman Alphabet at a Glance

#69
post #22

Earlier quoted context omitted.

No diacritics, and generous use of apostrophes.

> No diacritics An understandably naïve view of things whose errancy could easily be corrected through your coöperative perusal of The New Yorker.

Wouldn't it be cöoperative?

Re: Identify Any Language Written in the Roman Alphabet at a Glance

#70

There is a character only used in Taiwanese, not Chinese: 互 By that, I mean the Taiwanese language, which is not the same as Mandarin Chinese. Both languages are used in Taiwan, although Mandarin is the official language of the (outgoing) KMT government. Taiwan number 1 ;-)

What? Chinese and Japanese both use 互.
Post reply on HN