Wrong search results with Chinese labels

[Felix Grund (Scandio)]

Hi everybody,

One of our customers is experiencing a weird search issue in their Chinese department. I broke the issue down to the most simple case: one page has the label 其他产品 (other products) and one has the label 产品信息 (product information). When I do a label search like labelText:其他产品 both pages are found:

screen-pocketsearch.png

Does anyone have a clue why this happens?

Regards, Felix [Scandio]

[Stephen Deutsch · Accepted solution]

Hi Felix,

Taking a look at the documentation for the tokenizer for Lucene that deals with CJK characters, it seems like it splits up the characters into two-character bundles:

https://lucene.apache.org/core/3_5_0/api/all/org/apache/lucene/analysis/cjk/CJKTokenizer.html

That would explain why it matches the two characters for "product" in both strings.

[Felix Grund (Scandio)]

Hi Stephen,

Yes, we already solved this issue with Atlassian support. It actually works if you put the strings in double quotes. I'll accept your answer .

Regards, Felix