Zia Approval / Request Reopen prediction fails due to the presence of Email Classification Headers
Issue:
Zia Approval / Request Reopen prediction fails when the reply mails have Email Classification Header contents in them.
Disclaimer:
For eliminating Email Classification Headers while processing approval / reopen reply mails for Zia prediction, we already have a fix. Refer SD-108324. This fix was released in ServiceDesk Plus 14202 itself.
Solution proposed in that fix:
In order to discard such Email Classification Header contents, we are maintaining an outlookHandler.txt file under the directory: \ServiceDesk\ZIA\resources\signature_removal. Users will have to add the Email Classification Header content that they are using in their email manually to this file. Each entry in this file must be separated by a new line. For example, if the email classification header content is as shown in the image below (testemail1.png), kindly add "This Email is classified as: Important" entry to the outlookHandler.txt file.
Note:
1. After adding entries to the outlookHandler.txt file, make sure that you restart the ServiceDesk Plus application to reflect the changes.
2. If you want to add another entry, make sure there is a line break at the end of the previous entry.
3. You can not add non-English characters to this file, as Zia Approval / Request reopen prediction does not support other languages (till the latest release as of 21/11/2023 = v14505).
4. Since the outlookHandler.txt file can read contents only in English, even if you are using non-English characters in your email classification headers, kindly trim the email header to the last valid character (excluding white spaces) and then add it to the outlookHandler.txt file. For Example, if the email classification header content is as shown in the image below (testemail2.png), kindly add "Public -" entry to the outlookHandler.txt file.
5. From 14700 (Zia multilingual approval prediction feature release), this method is applicable only for Request reopening prediction
For CXs above 14700 and Approval Prediction:
1. Go to <Installation Directory>\ServiceDesk\ZIA\dataset folder.
2. Open email_labels.txt file with any text editor like Notepad.
Note: Take a backup of the file outside the Installation directory before making any modifications.
3. Keep the cursor at the third line because the file already has two default header lines.
4. Paste the email header content and save it.
New to ADSelfService Plus?
Related Articles
Zia prediction failure due to security statements at the front of email content
Issue: Zia approval prediction fails when security statements are present at the starting of the mail content. Since we consider the first 100 characters and 2 sentences from the content, the approval comment given by the user was not considered due ...Zia Prediction Fails When Email Content has Classifications - Extended Version
Issue : Zia Approval prediction is not working when the prediction query has email classification content in it. Fix : In this fix, we have created a file called IgnoreList.txt. The fix will read the contents present in this file and ignore them ...Zia Prediction Fails When Email Content has Classifications
Issue: Zia Approval prediction is not working when the prediction query has email classification content in it. Fix: In this fix, we have created a file called IgnoreList.txt. The fix will read the contents present in this file and ignore them while ...ZIA: Signature removal Issue in builds >=14202 (XGBoost algorithm)
Issue: Zia Reopen prediction fails because it is predicting the signature content of an email rather than the description. Fix: The greetings (hi, hello, dear, etc) and email labels (ServiceDesk\ZIA\resources\signature_removal\outlookHandler.txt) are ...Zia approval prediction failure even with customized data
Zia approval prediction is not working as expected even though the data is given under customized data. Fix: The customized model will be considered for the prediction at first and then if it fails to predict, the default model will be considered. ...