As I browse, I compulsively select paragraphs of text. I imagine that this habit could cause some interesting data problems, if done on a site which tries to track such actions.

Does anyone else highlight as they read without searching, i.e. how valid is their assumption: We all know, or at least suspect, that people on our sites are highlighting text, right clicking on it and selecting "Search Google For..."

I do that ALL the time, and I too was thinking that it would give you a bunch of garbage data.

But then I was also thinking, it could give you some useful measure of where people's eyes happened to be on the page- which copy they were reading and which they were skipping over.

I'm not the only one! I didn't think about users searching for text, but I did think that this could be useful to learn how much users are interacting with my site. (ie. Are they losing interest half way down a page? Are they drawn away from the main content body? Etc.)

This is smart. Wish I'd thought of this. I love hacking GA to do fun stuff. My recent favourite is tracking real page load time (i.e. not just server response times but actual user load times in the wild).

Perhaps you can give us an example on how to implement this. I would be really interested in it.

Any chance you could share that knowledge? That sounds fantastically awesome!

How about, when someone highlights text on your site, it immediately searches itself and presents the results in the margin? (Without causing the page to reflow or otherwise getting in the way.) Perhaps with links to search Google/Wikipedia/etc as well? Too creepy?

nytimes.com used to do something like this. It was super annoying.

That stuff is absolute hell for people who compulsively highlight the text they're reading. So are all the sites with links everywhere that pop up a little social networking/other articles windows on highlight.

I use a FF extension called FastestFox which takes highlighted text and automagically pulls a snippet from wikipedia and/or the dictionary (it is smart enough to know which) in addition to offering the option of searching it in one of several configurable search engines. Very helpful when browsing. And works way better than when the NYTimes tried it.

My MSc thesis was on exactly this - adaptive web sites. It's a fascinating research topic.

Did you ever implement anything that we could play with?

the new york times has had something similar to this for years.

you can highlight words and it will show a "question box" that you can click to see more information about the word, google search, etc. it's displayed inline so you don't need to navigate away.

it's a nice idea, but it's terribly annoying.

sadly not. I had it running on my own blog for a few weeks, collecting data which I then analysed. I'd have loved to have got a nice plugin put together for it all but I didn't get it that far.

Got a distinction for it though! You can read a little about it here: http://www.puremango.co.uk/2009/06/adaptive-web-sites/

It's awesome is what it is. I wonder if you can pull the 1,2 and 3 word terms via the Google Analytics API? I'd betempted to use that data to generate popular search pages / navigation and internal link changes.