Re: Can you help me to merge your avatars?

Le mardi 23 novembre 2010 à 15:00 +0100, Mathieu Goeminne a écrit :
> Hello there, 
> First of all, thanks for your work!
> I send you a mail because I asked a question on the Brasero IRC chat
> which said to ask it to you.
> I'm Mathieu Goeminne, a phd student (at the UMONS, Belgium) and I'm
> studying the free software ecosystems. A part of my job consists in a
> merge of all persons involved in free software (including Brasero and
> Evince). I designed some new algorithms, and implemented others to try
> to detect real physical persons behind a mail adress, a svn login, a
> pseudo, etc. In order to validate my algorithms, I need a "merge
> reference", ie, a list of correspondences between all evolved people.
> I firstly did it manually but there are still errors on my reference. 
> Have you a correspondency table such that using it I can say "this
> person and this one are actually the same physical person"? I'm
> focused on the committers, mailers and bug reporters.
I think that as Johannes said, there is no such list. You have a git
account, and Bugzilla account, and a mail account, but you're not forced
at all to use this e-mail address for the two other accounts.

> If not, have you an idea about a mean to build this kind of table?
I'm afraid you cannot, at least if you mean to get an exact information.
People that wanted this information just did like you, by using a few
tricks to approach the reality, but we can't be sure their succeeded.
See for example the GNOME Census, which has required some reflections
about this problem:

The only solution I can see is to validate a part of your data by hand,
and then get a measure of what your algorithm got wrong. Then,
statistical tests will give you confidence intervals that you'll be able
to apply to the broader survey. But I'm sure you thought about this
already. ;-)


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]