Wot's...Uh The Deal....
- 1 August, 2:09am |
- jaws
At this time I'm just seeing numbers and name of Bayes everywhere in my house. The reason is that I've been coding many different stuff for Jaws 0.6, one of the last hacks is about a Spam Filter.
Yes, many of us have been attacked by spam in our comments and chatbox, the most common spam message is about a 'Casino', you can see many of this spam messages in some entries of ion's blog (here's one). So I decided one month ago to write/implement some existant filter for Jaws and I decided to use one that is based on the Bayes theorem (or in spanish).
I decided to take a look in Google and find some existant Bayesian scripts and hooray! I found one in PHP so I started testing it and yesterday I considered it stable to port it to Jaws.
Now I'm glad to have a spam filter in Jaws for comments and entries on Chatbox, but what's missing?
- A nice UI to train the filter, I'm thinking in using a Popfile-like UI.
- Find more spam message examples. Yes! the installer now has a stage that adds like 3000 words that are classified as spam I found these words in Google.
- Think in writing more features for it.
I've been testing/training the Spam filter in the last 30minutes and now 90% of the messages are classified without any problems, some of them under spam and some others under the nospam category.
Here you can find a nice article that explains the relation between Algorithms and Spam.
In other hacks, I've been hacking the REST webservices for many gadgets, my favorite one is the Filebrowser REST webservice (to share directories with your friends), also I've been playing with the FastURLMappings and other stuff 
.
The last week I sent an email to the Eluniversal webmaster suggesting him to add LiveFeeds and RSS for Mozilla users, then he told me he will implement that... now that I'm reading the site I'm so fucking happy to see that Mozilla users can use LiveFeeds, one hooray for them!
