Environment
GWAVA 4 (all builds) on Netware and Linux
Situation
I've put ham in my spam corpus or vice versa
Resolution
There are two main events that occur when a message is put into your spam or ham corpus. One, the original message is stored in the ham/spam corpus. Two, the message is chopped up, parsed and message tokens are entered into the dictionary files. The dictionary files are used by GWAVA during spam scanning and are what determines whether a message is spam or ham.
Messages can be removed from the corpus, but the message's token will still remain within the dictionary files. As there is no way to manually clean out entries from the dictionary files, you are left with two options:
1) Eventually, the dictionary files and corpus are pruned and old data is removed, so over time the problem will resolve itself. This is the recommended solution, as long as the number of messages that were inappropriately categorized as ham or spam is relatively small in comparison to the total corpus. Note: an handful of messages isn't going to have a huge affect on your overall performance. In most systems, a few hundred messages isn't going to have a big affect.
or
2) You have the option of scrapping your current corpus and dictionary files and starting over. Here is a good article that covers this topic: https://support.microfocus.com/kb/doc.php?id=7019711 . Note: This article describes putting your dictionary files back to default. Rather than deleting all of your corpus, if you know there are good messages in the corpus that you'd like to reuse, then you can copy those messages from ...GWAVA4\services\autoblocker\[scanner id]\corpus\ham and spam directories back into your newly created dictionary files by putting the files in the ...\GWAVA4\service\autoblocker\[scanner id]\transfer\ham and spam directories. Sorting out which messages are categorized appropriately and which aren't will be difficult, but if you know that a certain date range is good, then you can copy only those messages.
Additional Information
This article was originally published in the GWAVA knowledgebase as article ID 308.