iFile finally classified something as belonging on my weblog, but I have no idea why… Justin Rudd’s Busy weekend complicated by O/R Mappers doesn’t sound at all like my weekend.
Tag Archive for 'bayesian'
Bayesian Aggregation, Part II
23Feb03CNET interviews Thomas Bayes
18Feb03Not really, he’s been dead for over 200 years, but CNET has published a longish article that takes a Microsoft-centric look at the applications of Bayes Theorem. Bayesian classification is useful for much more than Spam fighting…
Bayesian Aggregation, Part I
14Feb03On Monday I configured Scenario 3 of my Bayesian Aggregation experiment, building a “good” corpus of my weblog entries and a “bad” corpus of Sunday’s RSS items.
There are no results to report so far, and the only conclusion to be made is that I don’t know enough about iFile’s parser and classifier. Is my […]
False Positives
14Feb03On Sunday night I made a dramatic change to my POPFile configuration, removing 14 mailing list “buckets” — three weren’t accurate, the other eleven were simply not being used for filtering. I kept three mailing list bucks that were highly accurate (> 98%), plus Spam and Normal.
The immediate effect of this change was that Spam […]
Applying Bayes to Aggregation
09Feb03Last night I went ahead with the rss2email switch, splitting my subscriptions across several instances as a work-around for the problems I had. I’ve tweaked rss2email.py to create some debug output, wrapped all of the instances in a shell script that emails the output to me, and set up a cron job to run four […]
Spammer Tricks
02Dec02John Graham has posted about some recent spammer tricks that he is working to address in the next release of POPFile.
I’ve been very happy with POPFile as a spam filter, I’ve had zero false positives and very few false negatives. It hasn’t worked so well on my mailing lists, messages from radio-dev, rss-dev, and livetopics-support […]
In blogdex land, I was the first to link to POPFile. The TechTV coverage brought in a minor rush of users and posts to the POPFile mailing list. Slashdot’s coverage resulted in the usual flood. I can’t take any credit for the publicity, and I kick myself for not thinking to submit the story to […]
