Pages: 1
RSS
Bayes-It, Please document Bayes-It product
 
Hi,

I am a long time user of The Bat!.

I've just paid for the v3 version. Mostly to trash my popfile antispam solution in order to get Bayes-It integrated with the new version.

PLEASE : Give us a (non-obsolete) DOCUMENTATION on how to install, configure and start with Bayes-It Anti-Spam.

I've upgraded my TheBat from 2 to 3. I've installed and activated the Bayes Plugin (0.7.3, by this time).

I did mark as junk all my email since Friday (I did click on "junk" 427 times).

All parameters were left as default.

Bayes-It did'nt move any bad mail by itself.

I'm getting more than 200 bad mails a day. It's urgent :(

PS : At install time, after restarting, I've had message boxes saying the "my .lst files were bad". I've had to dig a lot, a found a lstrepair.exe that I did suceed in applying. Are my problems related to this ?
 
Here is my "Information" panel (see Error=100%):


----

BayesIt! filter information

Antispam filtering data:

Spam frequency dictionary:
° C:\Program Files\The Bat!\MAIL\BayesIt\base\spamdict.bye
° Size: 365 letters.
° Capacity: 27460 words.

Non-spam frequency dictionary:
° C:\Program Files\The Bat!\MAIL\BayesIt\base\nspamdict.bye
° Size: 49424 letters.
° Capacity: 1621896 words.

Current active base:
° Active current base contains 306999 words.
° Status: OK

Last day statistic
General numbers
° Spam traffic (bytes): 717447
° Spam letters: 121
° NON-spam traffic (bytes): 2000946
° NON-spam letters: 124
° Total traffic (bytes): 2718393
° Total letters: 245

° Part of the spam in terms of letters: 49.39%.
° []
° []
°
° Part of the spam in terms of traffic (bytes): 26.39%.
° []
° []
°

Errors
° SPAM errors (letters): 100%.
° []
°
° SPAM errors (traffic): 100%.
° []
°
° NON-spam errors (letters): 0%.
° []
°
° NON-spam errors (traffic): 0%.
° []
°
° Totally errors (letters): 49.39%.
° []
° []
°
° Totally errors (traffic): 26.39%.
° []
° []
°

Last week statistic
General numbers
° Spam traffic (bytes): 1985530
° Spam letters: 362
° NON-spam traffic (bytes): 58615427
° NON-spam letters: 1519
° Total traffic (bytes): 60600957
° Total letters: 1881

° Part of the spam in terms of letters: 19.25%.
° []
° []
°
° Part of the spam in terms of traffic (bytes): 3.28%.
° []
° []
°

Errors
° SPAM errors (letters): 100%.
° []
°
° SPAM errors (traffic): 100%.
° []
°
° NON-spam errors (letters): 0%.
° []
°
° NON-spam errors (traffic): 0%.
° []
°
° Totally errors (letters): 19.25%.
° []
° []
°
° Totally errors (traffic): 3.28%.
° []
° []
°

Last month statistic
General numbers
° Spam traffic (bytes): 1985530
° Spam letters: 362
° NON-spam traffic (bytes): 185611187
° NON-spam letters: 7002
° Total traffic (bytes): 187596717
° Total letters: 7364

° Part of the spam in terms of letters: 4.92%.
° []
° []
°
° Part of the spam in terms of traffic (bytes): 1.06%.
° []
° []
°

Errors
° SPAM errors (letters): 100%.
° []
°
° SPAM errors (traffic): 100%.
° []
°
° NON-spam errors (letters): 0%.
° []
°
° NON-spam errors (traffic): 0%.
° []
°
° Totally errors (letters): 4.92%.
° []
° []
°
° Totally errors (traffic): 1.06%.
° []
° []
°

Plugin exports the macros:
1. %BAYESBASE - returns the size of the current active base;
2. %SPAMINESS('word') - returns the grade to 'word' from the base.
3. %BAYESITVERSION - returns the name and version of BayesIt, like 'BayesIt! 0.5.3'.
4. %REGEXED - invokes visual editor for regular expressions, inserts the ready expression as a result of the macro
5. %STATSPAMLETTERS(hours)
5. %STATSPAMTRAFFIC(hours)
5. %STATNONSPAMLETTERS(hours)
5. %STATNONSPAMTRAFFIC(hours)
5. %STATSPAMERLETTERS(hours)
5. %STATSPAMERTRAFFIC(hours)
5. %STATNONSPAMERLETTERS(hours)
5. %STATNONSPAMERTRAFFIC(hours) - returns statistic for last x hours.


Plugin is written by Alexey Vinogradov (antispam@ritlabs.com)
Information and updates at http://www.ritlabs.com/ru/solutions/BayesIt.php
Current running version is 0.7.3
 
Oliver:

Not sure what to make of your information panel, however, I just started using BAYESIT yesterday and have junked only 190 emails to date.

I will wait to see what happens.

Did you get any feedback from anyone regarding how long one is supposed to have to wait to see any results??

kalaZano
 
No, I didn't get any answer from nobody :(

It's a strange feeling for a returning CUSTOMER :(  :(  
 
I agree totally with Olivier. I just upgraded to V3 and I feel ripped off. First of all the few small changes made don't justify a major version number change and upgrade, and secondly there is NO DOCUMENTATION for the BayesIt filter. Not just bad documentation -- I'm used to that from TheBat!, the documentation is always bad -- but NO DOCUMENTATION.

There is also not even any information on the website about how to use and configure BayesIt -- after searching for a long time I found the old information on http://klirik.narod.ru/usefuls/bayesiteng.htm, but this doesn't apply at all to the current version.

I've been using TheBat! for years and it is the best email client around but their documentation is really, really, really bad. I'm a technical writer and my job is to write help for Windows programs, and the documentation of TheBat! is pretty much the worst I have ever seen anywhere. I don't think there's any documentation mistake they left out -- when I teach classes and give lectures on how to write good help I always use the help file from TheBat! to show students how NOT to write help.  
 
I agree, this version of the bat is the first version that i have had any problems, and it is all because of the Bayesit filter.
I may have to go back to my other filter which i was not happy with, but worked ok.
 
Shortly speaking, the bayesit's info page is now under construction. Since I am the only man who do this, please, keep patience. I have to write about 50k or text, and then translate it into english. And I also have to do my every-day job, since I myself have nothing from daily development of bayesit.
 
Quote
Shortly speaking, the bayesit's info page is now under construction. Since I am the only man who do this, please, keep patience. I have to write about 50k or text, and then translate it into english. And I also have to do my every-day job, since I myself have nothing from daily development of bayesit.

Then you shouldn't have published the program because it wasn't finished. A program of this type without documentation is just as non-functional as a program with defective code. To work, the program must interface with both the computer and the user. Leaving out the documentation is just the same as leaving out API calls or part of the code -- it means you have published a broken program.

And why are you giving your work to RITLabs for free? I'm sure they are very happy to be making a lot of money off your generosity but you don't seem to be happy with it so why are you doing it??  :o  
 
The whole set of russian-language documentation is about to be finished tomorrow (I have 2 more "advanced technical" sections to write). Then I'll begin to translate it into english.

The documentation will include two documents - for simple initial setup, and for advanced users. Initial documentation describes the basic step in installation and initial setup of the module. Advanced documentation describes more complex questions like working with "selective download", Visual regexp debugger and a bit of explanation of some internal files and processes.

I can begin to publish the english version of initial setup as fast as possible, but I don't think that the official publishing of the "raw" (i.e., unedited and full of typerrs) document will be proper.

So, if you want, I can send this "raw" variant exactly here, in this thread of this forum.
 
Ok.Please do publish it first here.

I just hope that will solve my problem and BayesIt will move spam into the trash (or special folder) by itself.

Thanx
 

I see that Alex has not yet posted his english draft
translation of the BayesIt filter. I too, like most
users, are wondering how this thing works. I installed
it, but I don't even know how to get the thing started,
and get it trained to recognize spam. I can't find
anything in the menus and all I know is none
of my spam is being filtered. So if anyone can make
heads or tails of this crude translation, here's a
link to a translated version of the Russian BayesIt
documentation (I figure its better than nothing!):


http://www.online-translator.com/url/tran_url.asp?lang=en&url=http%3A%2F%2Fwww.ritlabs.com%2Fru%2Fsolutions%2FBayesIt.php&direction=re&template=General&cp1=NO&cp2=NO&autotranslat­e=on&transliterate=on&psubmit2.x=41&psubmit2.y=10
 
The way it is supposed to be trained is by marking spam as junk and marking ham as not-junk. Once you have built up a dictionary by marking "sufficient mail" it starts to recognise and distinguish more accurately for itself.

I personally use and prefer Achim Winkler's BayesFilter plugin. I have never managed to get the BayesIt filter to work well for me. Many others have done so just fine.

The nice thing about plug-ins is that there's always another one around the corner ;)
iviarck
 
A very complete Tutorial about bayesit can be found here:

http://www.thebatworld.de/system/sect...e&artid=10

The only drawback, it is in german. But maybe you can use babelfish.altavista.com to translate it.

It also tells you to rightclick on the plugin unter preferences to reveal an information tab. here you can see how you trained the plugin so far!

hope this is of any help

regards

moongate
 
I've finally written the russian manual for BayesIt. Now I have to translate it into English and thish is quite hard task - the text only is about 100kb, and I also have to make english skinshots as illustrations. I think, I'll do it by small portions, and will publish it here with the skinshots as the simple links. The "beginner's" part is quite short (34kb), and will take two or more days. The "advanced" part is much longer (more than 60kb), and I afraid that it may not be translated at all (I don't feel so much enthusiasm in myself for such task, and except this private enthusiasm I have no support from anybody (firstly, from RIT S.R.L) at all).
 
I think there's a bigger problem than documentation. I've now got 244 spam messages in my junk folder, all of which I had to put their manually, and I've marked thousands of emails as not spam.

Even dropping the figure next to "Move a message to the Junk folder if the score is greater than..." down to 50 and still it hasn't marked as spam a SINGLE email on its own.

To be honest, I don't think this plug-in even works.
 
For some it works. For me it never worked. Please investigate using the alternative plugin (BayesFilter) if BayesIt doesn't work for you.

See Achim's website for more information.
iviarck
 
I concurr, kick the .6 in the bucket and get Achim's 1.5.6 bayesit plugin.

Nicer, better!

Just having to re train = pain in the you know what.
 
Ah great thanks. Retraining is easy since I can just mark all the mail in my junk folder as junk again and mark the thousands of mails in my mailing list folders as not junk.

So far it looks like it's working too although I did need to tell it manually where the database files were. It would train but every time I checked the config it warned that it hadn't been trained. At least that allowed me to work out what was going wrong.

How about The Bat ships with this instead? You know, because it works and stuff.
 
Hi there,

just wanting to tell you guyz that i ditched the original bayesit plugin and installed Achim's and it works great !

thanks to him for that

R a d
 
Hi, I'm not sure where to post this but ..

I'm using the Bat 3.0.1.33 and Achim's Bayes filter. It appears to work but when I click special/mark as junk the program scans but the message doesn't get added to the blacklist. Messages marked "not junk" do get added to a white list. Any help would be appreciated.

thanks

MIke
 
Quote
Michael Redbourn wrote:
when I click special/mark as junk the program scans but the message doesn't get added to the blacklist. Messages marked "not junk" do get added to a white list.

Messages marked as junk should be scanned, the content assimilated into the "intelligence" database to enhance scoring and move to the designated Junk folder. They should not be blacklisted in any way. This is a transparent operation and is part of the "training" operation required by Bayesian spam detection systems.

The automatic adding of non-spam messages to the whitelist is an option in earlier versions of BayesIt and has been removed in later versions. It should not be used. Again, it is a matter of "training" the internal Bayesian detection logic and keyword database to be more forgiving.

Whitelisting of addresses should truly be a manual operation. As should blacklisting. Spam seldom comes from real addresses and even more seldom from the same address twice. Spam can also come from previously known addresses and blacklisting a valid message sender just because their address was abused by a spammer once is taking things too far.
iviarck
 
thanks. I got an email from the programmer and he told me to check both boxes next to "auto insert spam sender". I didn't have the training one checked. It's working now. He also said that I shouldn't check the box for whitelist learning. I have version 2.0.0 which I thought was the latest ? Regards, Mike
 
Oops - thought we were talking about BayesIt. Yes, Bayes Filter does have those options in all versions. Sorry for the (minor) confusion.
iviarck
Pages: 1