2.4 KiB
| title | date | techne | episteme | slug |
|---|---|---|---|---|
| Google Web History | 2012-02-27 | :done | :personal | 2012/02/27/google-web-history/ |
Since 2008-03-04 Google tracks all my searches. Thanks to [XiXiDu][xixidu search], I noticed that I could download and analyze the whole data set. There's an existing script, but it didn't work for me, so I wrote my own. It's up on [Github][github web history] and should be fairly self-explanatory for Ruby users.
Anyway, instead of spamming Twitter, here's some interesting results.
The history spans ~1.4k days and has ~132k entries, of which ~37k are searches. The rest are mostly search results I clicked. That makes it ~25 searches per day. Not bad.
My Top 10 search terms, ignoring word order:
- ezt (23)
- read (23)
- fefe (18)
- pharyngula (15)
- reader (15)
- 吉井和哉 (13)
- nanoc (12)
- muflax (11)
- squid314 (11)
- dexter (10)
Everything except 6), 7) and 10) are failed attempts to enter a (partial) URL and pressing return before Firefox could autocomplete it. (Yeah, I once read Pharyngula. I know.) 6) is one of my favorite musicians, but I'm quite surprised I googled him that much. 7) is the blog engine my main site uses, so no surprise. 10) is probably also an URL, maybe the episode list or something.
The terms are not surprising, but the low frequencies are. Maybe Google misses a lot of terms due to me not being logged in at the time, but I never consciously log out, so still weird.
More results:
- I google a lot of TV shows.
- I internet-stalk way too many people. I gotta stop that. (Not likely.)
- I google Drew Carey more than porn. And I thought I liked Ryan more.
- [Aki Sora][] is the porn I google most. Well, since my favorite sites shut down anyway and I have to hunt down my hentai like a damn savage again.
- Buddhism-related searches and "how the fuck does this totally normal item work" occur about equally often.
- I only once searched for Jesus. Gotta give the guy a second chance.
- I have googled "google" 3 times. I regret nothing.
- Automatic completion is treated like a shortened search, thus "read" and "reader". This probably hides a lot of searches and might explain the low frequencies.
And finally, number of searches for each hour:
<%= image("hours1.png", "hours") %>
Well, shit. You can totally [reconstruct my sleep cycle][Gwern anonymity] from that. Totally thought it would be more chaotic. Guess I really do have an underlying sleep cycle after all.