Re: [edict-jmdict] aruyouni

["Paul Blay" <blay.paul@**************>]

> Ummm...  Okay?
>
> So when *you* give the list of estimates as 30000, 141000, 12200,
> 12200 and tell me to spot the odd one out (in this case, 141000) in
> order to demonstrate that I'm incorrect, and then *I* give the list
> of estimates as 22500, 147000, 147000, 150000, you tell me that the
> odd one out (22500) somehow proves that I'm incorrect...
>
> Isn't this called "trying to have it both ways"?

The fact that you apparently get different results for the same
links only goes to show that something screwy is going on
(which should already have been obvious).  My position is that
when something screwy goes on with Google numbers it, practically without
exception, produces overestimates not underestimates.

(The one exception being when it thinks you are searching on a Chinese
search string).

The first thing to consider is /what/ in particular is screwy this
time around.  One thing that Google does is treat certain words
as if they were identical (ダイヤモンド / ダイアモンド)

If I search for 蛙 in Japanese language pages only (otherwise it will
think it is Chinese) I get 666,000 as the estimate.

カエル in Japanese language pages returns 2,260,000.  As it is well
known that 蛙 is a (uk) this result is plausible.

If I take the link that _did_ get me 141,000
http://www.google.com/search?hl=en&q=%22%E8%9B%99%E3%81%AE%E3%82%88%E3%81%86%E3%81%AB%22&btnG=Search
and change it to カエルのように
I get a result of 15,800
http://www.google.com/search?hl=en&q=%22%E3%82%AB%E3%82%A8%E3%83%AB%E3%81%AE%E3%82%88%E3%81%86%E3%81%AB%22&btnG=Search

So you have the bizarre result that a (uk) word is found about 10 times as
often in kanji than in kana.

Seriously, any time* there's a choice of "small result" or "large
result" it's the large one that's stuffed up.

* Excepting pseudo-Chinese text search