Friday, September 5, 2003, 10:04:49 AM, MAU wrote: >> How about applying Bayesian logic to this? Can filters call >> an external program or a plug-in?
M> Yes, on the Actions tab of filter you can select "Run external M> program". About Bayesian logic, I don't really know but I M> wouldn't think so. What is needed is a tool that would "learn" what you consider relevant and then process and score each individual e-mail. The Bayesian filters (I use POPFile) are moving in that directions, but are are not really sufficient. As a sci-fi reader, I await the day when the list of articles I'm interested in appears each day on my desktop! A possibility given current technical capabilities is to recognize certain patterns that reflect poor list "manners" such as: Too great a quoted text to original text ratio Top posting Etc. Then those could be detected and scored similar to what I suggested above. M> What we could perhaps try is to invent a Relevant Text M> Delimiter, much like the "-- " signature delimiter, and ask M> everyone to use it to mark where the real content (if any:) of M> their messages starts. What would you think of "%% "? Should M> we propose it? Wouldn't work. The people who would know about and follow putting "%%" as a marker already trim quotes, etc. adequately. Training people to do extra to make it easier for others doesn't work. Just note all the posts by the mods on this relevantly technical list just to get people to put sig delimiters in place and to not top post. M> Sorry, I'm not in TBOT. I'm going to still cross-post this to TBOT. Otherwise, I'm going to be wearing a big fish target across my chest. -- Dave Kennedy ________________________________________________ Current version is 2.00 | "Using TBUDL" information: http://www.silverstones.com/thebat/TBUDLInfo.html
