Duplicate AI compares tickets and will categorize them into duplicates and similar tickets. These are shown in seperate sections:
The distinction is made, since ..
.. for duplicates we are really sure you want to something about them.
.. similar tickets are just interesting.
Similar Issues
These are easier to explain, so we start here. To find them we calculate several similarity metrics (e.g. TFIDF, Levenshtein Distance, ..).
Those with high scores show up in the "Similar Tickets" section. To make it easy to spot interesting tickets, we show:
Ticket Summary
Ticket Status
A similarity score in % to indicate how far appart the similar ticket and your open ticket are
Duplicates
These tickets have passed several checks regarding their similiarity metrics and ticket properties. These checks on a combination of properties & metrics ensures that these are real duplicates.
Each duplicates also is assigned a "scenario", which indicates how we think what type of duplicate this is (e.g. 100% same content, email that references an existing ticket, etc).
Example - Scenario "Email followup"
Description is not that similar
(since in an email thread people add completely new content in the next email)
Summary is very similar
(since the email subject is usually the same)