For your last point, you can probably check with your Chinese friend and I'm pretty sure they don't buy the idea. Basically our brain already has built-in text

For your last point, you can probably check with your Chinese friend and I'm pretty sure they don't buy the idea. Basically our brain already has built-in text segmentation (https://en.wikipedia.org/wiki/Text_segmentation https://en.wikipedia.org/wiki/Text_segmentation). I tried to think how I actually do that. It's more like prefix matching over a moderate large dict. It's not complicated and worked pretty well for Chinese. For your particular example, maybe the difference lies in word lengths. Most Chinese words are under 2~3 characters, makes prefix matching way more efficient.