Rated 5 out of 5 stars

Works really well, I just had to train it on a couple hundred emails then it tagged then right 95% of the time. (I trained it another 1000, but that did not affect its success rate)

Rated 2 out of 5 stars

First of all: I appreciate any attempt to make email classification more "smart" . Actually, there is nothing dumber than standard techniques like manually setting filters.

This said, I wonder why it seems to be so incredibly challenging to introduce such functionality.
I enjoyed the built-in M2 Email client of the Opera browser from 2004 to 2013 till they finally dropped support for Linux and all went down the drain.

The M2 client had a "simple" technique: For each folder you created, you could tick "learn from email". When now an email was dragged into a folder, it learned with simple ham-spam-like classification if a mail fit there or not.
In the beginning, a lot mails ended up in all folders, but by drag-and-drop you were able to train the classification easily. The classification-file was just a text-file where you could check for each folder, which keywords were created by the system.

17 years later, I still find myself in email-stoneage, and all I can find is unintuitive, half-ready ****, that is no fun at all.

Rated 4 out of 5 stars

Excellent add-on! After spending some time teaching it. it sorted email accurately enough!
I would give 5 starts if it came with 3 or 4 categories pre-taught like gmail does.

Rated 3 out of 5 stars

I really like this extension, but I think it could get much better, here some proposals:

- Interaction with filters:
AutoBucket seems to tag my mails _after_ my filters are applied. As a result, in the IMAP inbox the tags are applied, but only after my filters have copied the mails to the local folders. In my case, it would be useful that AutoBucket would apply before the filters, or that you could choose if after or before. It would be useful, for example, to be able to use the tags of AutoBucket to filter the mails.

Another issue seems to be that AutoBucket only applies to mails which did not get any tag from the filters. I like that, but maybe this could also be user defined. It would be useful for the learning that AutoBucket would learn from the tags applied by the filters.

- No points boundary: It would be very useful to be able to set a bounday below which no tag would be applied. Right now, _all_ mails get a tag from AutoBucket (except those already tagged by the filters), but often the mails should not get a tag. AutoBucket should tag only when the tag is clearly to be applied. For example, some mails get a tag with 0,009 points, that is nonsense. A boundary should be able to be set. I would like for example a 0.2 or 0.3.

- I don't know if this is only my issue, but the learning and judging through context menu is extremely slow.

- I also don't understand if re- or untagging false positives corrects the learning. And also don't understand if learning is done only manually or if each tagged mail is learned (and unlearned if the tag is corrected). Maybe a bit more information would make it easier to understand the function. And, as I said above, learning from filters would be great (maybe integrating an action in the filters, something like "learn as X in AutoBucket").

Rated 3 out of 5 stars

MAJOR IMPROVEMENTS IN V.1.2.0!!!!
.. but very far from being "done"...

Whatever the Autobucket change was from "Sync" storage to Local storage, Autobucket no longer blows up memory use/pagefile. I estimate that the memory resource reduction to be >>70% from previous releases that were alleged to be bugfixed. DO NOT REGRESS THIS MAJOR ARCHITECTURAL ERROR IN FUTURE RELEASES...

It's possible that Thunderbird v78.11.0 bugfixes might be playing a part in Autobucket's rehabilitation, too.

No way to estimate how much faster 1.2.0 is over previous releases, because previous releases would hang Thunderbird for more than 24hrs before regaining UI response after ANY single Autobucket operation. Many orders of magnitude faster now, but still too slow - too slow for anything other than non-production hour runs of batch operations..

Autobucket desperately needs an UNLEARN AS TAG flyout menu option, RATHER THAN THE WHOLESALE DELETION OF ALL ACCUMULATED PREVIOUS "LEARNING." This is necessary whenever Autobucket egregiously mis-tags messages, even after HUNDREDS of manual Learn as Tag "lessons." WRONG TAGS, PER MESSAGE, NEED TO BE GRANULARLY UNLEARNED as part of "learning," whiile preserving all other valid rules.

THIS IS MAJOR PROGRESS!!!

Now, Autobucket is a modestly useful app, instead of [CR]app.



_____________________________
v1.1.3 "update"...
1) AB Options tab now only populates with NO parameters (NO Tags & ZEROED LIMITS, no matter what might have been saved before)
2) no option to Learn message as during right click on message list item
3) still chews up and down on memory consumption/exhaustion as reported with v1.1.0
____________________


Allegedly v1.1.0 works with Tbird 78.x.

Well, here's how...:

1) one of the SLOWEST extensions to load/reload that I have ever witnessed

2) Options page loads EMPTY, hangs (literally for hours), "recovers," then refuses to accept any parameters (sees no existing Tbird tags, accepts no Learning upper limit size, accepts no Judgement log retention), then hangs again,

3) worst possible resource consumption of any extension I have ever witnessed: can variously demand 2GB->20GB of memory while DOING ABSOLUTELY NOTHING, just "idling"). Luckily, it's not a memory leak - all memory is freed up if/when Tbird is exited and/or process-killed

I could go on.

Suffice it to say that this is a really excellent extension in hanging/tying up all of Tbird via ridiculous system resource exhaustion..

(Tbird really needs to be able to throttle runaway resource consumption by badly coded extensions like this one).

Are any of these the reasons why v1.1.0 is BURIED on the Other Versions page??? It sure doesn't show up on the main page.

This user has a previous review of this add-on.

Rated 5 out of 5 stars

amazon tag add-ons, let email more clear :)

This review is for a previous version of the add-on (1.0.11). 

Rated 1 out of 5 stars

I really like the _idea_ of this add-on - using bayesian flitering to tag emails.

The problem is that I cannot figure out how to train emails.

Is it just tagging them? Or do I need to use the context menu? Does un-tagging correct a false positive?

Also, it'd be helpful if there were default settings for the plugin, instead of blank fields. How many kb? No idea. How long for retention? No idea.

I'm guessing that this is just a documentation and UI/UX problem, not a functional one. But since I cannot determine if things are working or not, I cannot use it.

Thanks for writing it, and I hope that a future revision has some more clarity. Or perhaps a happy user can leave a comment here?

This review is for a previous version of the add-on (1.0.7). 

Thanks for the review!

I added the explanation to "About this Add-on".
I also have a help page. I want you to see that too.
I recommend setting 100KB and 78 hours. I was supposed to set the default value, but it wasn't working because of a bug.

I fix ver1.0.8

* Fixed the problem that the initial value of the option was empty
* Added instructions on how to use when installing an extension
* Fixed help

Rated 4 out of 5 stars

Danke :-)
danke für die Übersetztung

This review is for a previous version of the add-on (0.9.9).  This user has a previous review of this add-on.