LPM: cgi script

Joe Hourcle oneiros at dcr.net
Wed Apr 19 07:12:20 CDT 2000



On Tue, 18 Apr 2000, Frank Price wrote:
> On Tue, Apr 18, 2000 at 05:28:11PM -0400, Joe Alexander wrote:

> > The idea is in three parts. First, collect from a finite number of
> > respondents the answers to 40-question multiple-choice survey, and store the
> > results in a database on a server. The second part is to provide the same
> > 40-question survey (over the net) to an unlimited number of respondents.
> > Thirdly, the results (by each question) then are automatically compared to
> > the pre-existing results, and the program generates the best match between
> > the live respondent and the pre-existing results. This match is then shown
> > to the live respondent: “Your answers most closely match so-and-so’s
> > answers.”
> 
> It doesn't sound too tough.  Probably the hardest part will be
> determining conceptually what constitutes a "best" match.  If you have
> a good idea about that, implementing it might not be too hard.  That's
> easy to say from here :-)

I'd have to agree on that one.

It's one thing to say 'this person has 38 of the 40 the same as you', and
it's another to have each question being a ranking from 1-10, and so, you
might do something wierd such as a sum of the difference of squares to
find the closest match.  (so 3 questions that are 1 off on the scale is
considered a closer match than 1 that's 2 off, or something of that sort)

And of course, there's the issue of what to do if you just happen to get
people from your original surveying who have the exact same set of
answers.

[however, what this is reminding me of is the 'which star wars character
are you most like' web page....I think it might have been at Brunching
Shuttlecocks  <http://www.brunching,com/>, as they tend to have stuff like
that.]

-----
Joe Hourcle




More information about the Lexington-pm mailing list