[Debtags-devel] AI-Tagging further steps
Thaddeus H. Black
t@b-tk.org
Fri, 29 Oct 2004 13:00:57 +0000
--/9DWx/yDrRhgMJTb
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
Ben writes,
> ~/lang/perl/autodebtag> ./bayesian-tagger.pl data__font
> Tested packages: 30
> Matches: 30 ^=3D 1
> Mismatches: 0 ^=3D 0
This tends to prove the Bayesian concept, doesn't it?
> ... Another thing to keep in mind
> is that false positives should be better than false negatives, as false
> positives will show irrelevant results to the user, but false negatives
> will hide will hide relevant packages from the user.
A general debtags remark or two, not especially in response to Ben's
words:
The very nature of the debtags project is to resist automation in its
core activity: tagging packages. Were it not so, Debian hackers would
surely have attacked the tagging problem years ago. If coding is fun
but tagging is boring, then, well, ... I don't quite know what to say.
What do you say? We cannot altogether code around this problem, I
think. The code is very important. It provides needed structure. It
automates automatable tasks. It helps a lot. But the code is only part
of the answer. The other part is as digging a mine in olden days. We
take our shovels. We heft our picks. We dig through the hill, five
meters a day.
(It has been observed that hackers do not like to dig through the hill,
five meters a day. We like giant diesel boring machines and dynamite,
don't we? Maybe so.)
I tend to accept Ben's judgment in favor of false positives, although
one hopes that the false positives are meant to be shown to a human
editor (who can eliminate them) not directly to the end user.
--=20
Thad
--/9DWx/yDrRhgMJTb
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org
iEYEARECAAYFAkGCPwkACgkQh3E0gzgBXn7xBACfaG1yrny7mzawZ29qjqXrrQJd
nG4AoL3cjHZAJLxWcFtNYw6PEQFh9WVX
=GYVO
-----END PGP SIGNATURE-----
--/9DWx/yDrRhgMJTb--