Hacking with Greg: SWIM-ing with Google

After blaming my alarm clock hack on my wife’s penchant for the snooze button at a few conferences around London last year. I’m blaming her again. Let me rephrase that, Ron if you are reading this you are my muse, see that’s better isn’t it? Anyway Ron never uses quoting in Google to group search terms. She will happily enter multiple search terms and then trawl through them until she finds what she wants. I’ve explained using quotes to her before, but for her It’s just easier to brute force it.

So I had two options, continue to pester her, which would probably be bad for me. Or try and figure out a way around it. Having a keen sense of self-preservation and an aversion to sleeping on the sofa, I chose the later.

Anyway, Perl has an acronym DWIM meaning Do What I Mean. So I wondered why Google couldn’t just SWIM (Search What I Mean), without the hints of the quotes.

So say I am searching for a book by Larry Wall that I remember was colored pink. Now a search such as,

Larry Wall pink

Would bring me back a lot of Pink Floyd fans called Larry talking about the band's most famous album. However I have a theory. People tend to type words that are grouped such as Larry Wall faster than other words, or rather they type the letter after the space faster than they would if it was not part of the same grouping. So a fairly dumb computer program could look at the typing space and group Larry and Wall together.

So to avoid the sofa and to try and see if my theory is correct, I give you SWIM.

Comments

Typing "Liz Williams chocolate" (without the quotes, obviously), I get "Hmmm, it looks like you used the cursor, backspace, mouse or something else too complicated for this proof of concept. sorry about that."

Posted by: Liz Williams | 2006.03.14 at 08:18 AM

It rocks Greg - good work!

Posted by: se71 | 2006.03.14 at 08:59 AM

Liz,

yip there seems to be a bug there, sorry about that. for a proof of concept i really didn't want to worry about cursor movements/deletions etc. i'll scratch my head over the weekend and see if i can come up with an easy way to do it.

Greg

Posted by: Greg McCarroll | 2006.03.14 at 09:09 AM

I couldn't get it to work in Safari, Greg. But in Firefox, it's great.

I did find that multiple sets of dual words were rolled into one set of quotation marks, eg what should have been "George W Bush" "bald eagle" "stetson hat" came out as "George W Bush bald eagle" "stetson hat". Or was it my typing?

Posted by: Peter Nye | 2006.03.14 at 09:41 AM

Peter,

I think thats just your typing ;-) Actually its hard to calibrate the timings, again another piece for the weekend.

Posted by: Greg McCarroll | 2006.03.15 at 12:21 PM

Maybe String::KeyboardDifference could be used to help.

It's what I used to make the automated typing for the libapi screencast a little more realistic

Posted by: Mark Fowler | 2006.03.15 at 12:45 PM

Nice!

Posted by: David Cantrell | 2006.03.15 at 06:04 PM

Did you mean String::KeyboardDistance, I couldn't find anything on KeyboardDifference...?

Posted by: Ben | 2006.03.17 at 07:43 PM

Although I did NOT use the cursor, backspace, mouse, etc ... no matter what I typed (one word, multiple words, different pause patterns), I got: "Hmmm, it looks like you used the cursor, backspace, mouse or something else too complicated for this proof of concept. sorry about that." Firefox 1.5.0.1 on Mac OS X 10.4.

Posted by: Kevin Murphy | 2006.03.23 at 09:46 PM