[Debtags-devel] AI Tagger

[Benjamin Mesing]

Hello

> > > Yes, really speaking about different vocabularies. There's the
> > > "official vocabulary" at people.d.o/~enrico - this is being used to
> > > sync unstable Packages list too.
> > >
> > > My point was: Is there enough code to let AI tagger and bayesian
> > > filters produce a second and maybe a third vocabulary ? Considering a
> > > interface (web? gui?) to edit these new vocabularies "manually". Do
> > > you see? The goal isn't kill the first and current vocabulary but just
> > > see how far the others can go and start merge the good things produced
> > > through those scripts into the "official vocabulary",  jmo.
> > You are talking about a tag database not a vocabulary i guess?
> 
> Yes, sure. The vocabulary should be the same, sorry.

Note, that my algorithm works only if it was trained with input data
(i.e. manually tagged database). So the AI-tagged database will be
always based on a manually tagged one. Enrico did suggest something
similar to your proposal, by creating a tag-patch with the AI-tagger.
The patch must be reviewed manually.
I think forking a new database is not worth the effort. First the AI (at
least my approach) has no sufficient precision. Additionally I don't
see, that anyone would be willing to edit another tag database to
increase the precision.
Sorry, but I don't see any real benefit with your approach. 


Greetings Ben