Is Chinese text on webpages already segmented?

[webmagnets]

When I go to an English language web page and long press on a word, it will highlight the entire word. I can understand how the browser or OS knows where the word starts and finishes because it has spaces.

 

However, with Chinese it still knows where the words are. When I long press on the 大 of 大家, the entire 大家 gets highlighted. How does that work?

[Publius]

Doesn't seem to work for me.

[wibr]

On iOS you see a word segmentation happening if you select something, but that's a function provided by iOS, not by the website.

[webmagnets]

I'm seeing this on my Chromebook, but not on my Android.

[imron]

I would guess it's realtime segmentation done by the OS.  It'll never have to segment much text because it only needs to look forward and back to the nearest punctuation and/or whitespace. 

[mikelove]

Can confirm this is the case, yes; AFAIK all major OSes now include some sort of Chinese word segmentation support, though not every browser / text editor necessarily taps into it.

 

The default approach (used by anybody without the AI chops to do better) is to use ICU's dictionary-based word segmenter, which finds possible breakdowns using a Chinese word list and then picks the most likely one based on word frequencies. (pretty much the same thing we do, though our dictionary's bigger because we're not asking OEMs to devote flash storage to it on a billion devices ?)