statistics

[elsmi]

According to the Google n-grams, the single, number one, most common English sentence (and the most common way for a sentence to start) is:

"All rights reserved"

This fact is, for many reasons, meaningless.

But it makes me feel sad.

Comments

[arkuat]

But I don't begin any of my sentences that way. And I don't think you actually converse f2f with anyone who begins their sentences that way. So I can only conclude that your sadness is at least as much about the assumptions made by Google n-grams sampling as about anything else.

[elsmi]

Oh, of course not -- that is one of the reasons that it is meaningless :-).

It isn't so much making assumptions, though, as being un-picky -- the n-grams sample exactly what they claim to, i.e., everything. Based on this, it seems... more plausible than not... that out of all English sentences ever produced[1], more of them really are "All rights reserved" than anything else.

[1] Not that "produced" here is particularly well defined either. Basically, it's a factoid. But such a *striking* one.

public web pages != the English language

[xoddam]

"The n-gram counts were generated from approximately 1 trillion word tokens of text from publicly accessible Web pages."

That's rather different from "all English sentences ever produced".

Re: public web pages != the English language

[elsmi]

Yeah, that's one reason why I had that "more plausible than not" hedge.

The web is a substantial fraction of all English sentences ever produced (a lot of those pages are computer-produced, and consider that computers can produce sentences a lot faster than humans otherwise do), and humans just don't produce the same sentences over and over (unless you count things like backchannel utterances -- "uh huh", "yeah", which perhaps you should). That's really a lot of what's going on, and what makes it both plausible and meaningless.