Moved from W10 to Linux (Rocky/RHEL 9) and spam filtering seems broken
I understand the adaptive methodology (I've coded Naive Bayes algorithms) but since the move from a Windows 10 workstation to a Linux workstation (three months ago), I've been marking a LOT of spam emails. I'm wondering how much data the algorithm needs or if there's a setting I need to change. I'm also wondering if the feature works the same on Linux and whether I can migrate the fitted model that worked well from the W10 TB profile (?) to the Linux TB profile. Also, I'm wondering if there's any way to get insight in what state the model's in.
I'm sure I can find more info by digging around, but I was wondering if anyone had any comments or clues or was ever in a similar situation. If there's a problem with junk filtering on Linux, it makes TB a lot less useful.
-Alan
Chosen solution
For your first question, I believe you can copy training.dat from one OS to another.
As for insight, the algorithm hasn't changed much in 15+ years, but not many tools to gain insight. But some references and tools do exist.
https://support.mozilla.org/en-US/kb/thunderbird-and-junk-spam-messages is one starting point. And a web search should find one or two standalone tools that can analyze the training.dat file.
https://addons.thunderbird.net/en-US/thunderbird/addon/junquilla/ is/was a nice tool, which perhaps you could adopt and update to work with a current version of Thunderbird. It offered user display and control to tweak Thunderbird's behavior.
Lastly, https://mzl.la/3HYAWos is a list of bug reports which mention "score" or "junquilla" which should offer some insight. And for broader perspective, https://mzl.la/3I4JmdT is a larger list of spam/junk bug reports.
Read this answer in context 👍 1All Replies (2)
Chosen Solution
For your first question, I believe you can copy training.dat from one OS to another.
As for insight, the algorithm hasn't changed much in 15+ years, but not many tools to gain insight. But some references and tools do exist.
https://support.mozilla.org/en-US/kb/thunderbird-and-junk-spam-messages is one starting point. And a web search should find one or two standalone tools that can analyze the training.dat file.
https://addons.thunderbird.net/en-US/thunderbird/addon/junquilla/ is/was a nice tool, which perhaps you could adopt and update to work with a current version of Thunderbird. It offered user display and control to tweak Thunderbird's behavior.
Lastly, https://mzl.la/3HYAWos is a list of bug reports which mention "score" or "junquilla" which should offer some insight. And for broader perspective, https://mzl.la/3I4JmdT is a larger list of spam/junk bug reports.
I marked this as solved because Wayne's answer is so helpful.
BTW, the specific issue is that TB is marking emails as junk but it's not moving them to the junk folder. I just reviewed my settings and I think I've configured TB to move mail determined to be junk into the junk folder, so that's why I'm wondering if there is a bug.
Modified