-
Suggestion
-
Resolution: Unresolved
-
1
-
5
-
Issue Summary
The new search experience doesn't take account of the index language.
Steps to Reproduce
- Change the Indexing Language config to CJK
- Create 3 pages which have a title 消費税, 消去 and 増税対応 accordingly.
- Go to the old experience and search pages with a query 消費税.
- Check the search result shows the pages containing 消費税 in their title or content body at the top.
- Go to the new experience and search pages with the same query
- Verify the search result contains the pages which title is 消去 and 増税対応 in higher search rank
- When the number of pages increased, the search result will be useless
The exact-match operator is unavailable with the new experience neither(CONFCLOUD-67993)
Expected Results
The pages corresponding to the search query will be ranked in the top.
Actual Results
Irrelevant pages are ranked in the top.
Workaround
Consider adding yourself as a watcher to be kept informed as to the state of the bug moving forward. With that way, if our development team updates the ticket, you'll be notified via email.
Bookmark the old experience and use it for the search.
https://EXAMPLE.atlassian.net/wiki/dosearchsite.action
Thanks to CONFCLOUD-67993, the exact match operator is now available:
With the CJK configuration, the search appears to behave differently up to the target:
page bodies: Kuromoji
page titles: N-gram (unigram)
- For Japanese-speaking users, Kuromoji is the best possible generic search engine.
- You maybe want to name the language configuration "Japanese" but not "CJK" because the name CJK seems suitable for N-grams but not for Kuromoji.
- The known workaround with the exact match works well for N-grams but not Kuromoji.
- relates to
-
AI-226 New advanced search returns irrelevant results for the exact match
-
- Closed
-
-
- Closed
- mentioned in
-
Page Failed to load
-
Page Failed to load
-
Page Failed to load
-
Page Loading...
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
| Form Name | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
[AI-852] The new search experience doesn't respect Indexing Language configuration
Hi everyone,
This is Aniket from the Confluence Engineering team. We would like to thank you for all the comments on the current need to target search relevance improvements by language. The current functionality of search relevance is expected as we are still working on enhancing our search systems to support relevancy for different languages. I have thus converted this Bug to a Suggestion so that your votes will carry over and you can continue to track our progress.
Best,
Aniket Bhosale
Hi everyone,
Thank you for previously raising this bug and bringing it to our attention.
Within our company roadmap and work capacity, we try to address or review each bug request but admit that not each one will be resolved. To continue the culture of being honest and open, we are closing this bug to focus on our upcoming roadmap for all Confluence users.
As we continue to roll out features we do look at requests made by our users and if you feel like this bug is still impacting your team please let us know.
Thank you again for providing valuable feedback to our team!
Hi all. Thank you for flagging and following up on this issue. We have a number of initiatives planned within search in the next few months that we expect should help with relevancy in character languages. Please stay tuned and thank you for your patience!