Detecting duplicates via "Your topic is similar to..."


#1

We recently made some changes so that the “your topic is similar to…” code uses title and body to try to detect duplicate topics as you type.

Here it is in action on this very post after I paused typing:

This may only apply to @leedohm and @AbeEstrada but we would like feedback on how well this feature is working. I know there are a lot of duplicate topics here and this is a difficult problem – at some level you can’t force users to read and if they are oblivious about creating a duplicate topic there’s no way to physically stop them – but I want to make sure our duplicate check panel is functioning.

So the next time you see (or create) a duplicate topic, try mock creating a new post and see if the Your topic is similar to… code would actually have shown the user a valid duplicate target?

I know duplicate topics is a big, tough problem. If there are any suggestions about ways we can improve this beyond the body check that @eviltrout added recently (prior to that the panel was based exclusively on post title), please let us know.


#2

I could have sworn I noticed that the panel updated itself after I’d been writing my post a few sentences and so I kinda assumed it *did * actually use the body. Good to know it will now though.

I think even as a thoughtful and caring member of the community it’s so hard to work out whether you are posting a duplicate or not. Obviously any way the system can help us out is great, but sometimes you see a topic that looks the same but when you check it out it’s either quite different or slightly different. Maybe this is a case where we can flag a topic for moderator attention (as per this reply) to get a subject line made clearer.


#3

Titles are hugely important. Discourse users who make it to trust level 3 can edit titles and categories on topics, for this very reason…