[Kde-pim] [akregator] Bayesian approach to ease news reading...

[Frank Osterfeld]

On Sunday 01 February 2009 16:17:40 jack wrote:

>  Using bogofilter [1] it would be "easy". When you read some new this
>  data is considered as ham, and the db is updated. All the news don't
>  read are considered as "spam".. then every time you read or not a new
>  you are teaching the algorithm. Bogofilter reads a message from stdin
>  and returns 0 (spam) or 1 (non-spam).. using that logic it could
>  changes the background color of the new.
>  That would be the initial idea.

I have no idea how such an approach it would work out for RSS feeds, but I find 
it a quite interesting area for experiments and research.
Some years ago, a student started to experiment with classification techniques 
such as bayesian filters and Latent semantic indexing. It didn't get very far 
though (and the code is not in SVN).
Actually, this was the original reason to add the (now removed) 
ArticleInterceptor interface I mentioned in the other thread ;)
If you want to implement bogofilter support, we can either re-add 
ArticleInterceptor in trunk and implement it as an akregator plugin, or 
implement as Akonadi agent based on krss (trunk/playground/pim/krss). 
The former has the advantage that you can integrate it directly in akregator 
as it is now and requires probably less ramp-up time than getting used to 
akonadi. krss, on the other hand, is the future backend, avoids later porting 
to akonadi and might be also more accessible when experimenting (One can hack 
anything he wants into the demo rss reader of krss without fear of breaking 
stuff). You can also add additional item properties such as bogofilter scores 
there more easily. As we'll have virtual search folders there, bogofilter 
results (like a search folder with all items above a certain treshold) would 
be even a good test case there.

-- 
Frank Osterfeld -- frank at kdab.net
Klarälvdalens Datakonsult AB, Platform-independent software solutions
_______________________________________________
KDE PIM mailing list kde-pim at kde.org
https://mail.kde.org/mailman/listinfo/kde-pim
KDE PIM home page at http://pim.kde.org/