900922 - Simple quote (') and apostrophe (’) don't cut words

Status: RESOLVED WONTFIX

(support.mozilla.org :: Search, defect, P2)

Description

[Scoobidiver (away)]

The apostrophe (’) sometimes written as simple quote (') doesn't separate two words leading to no results while there are some.
It's often used in French and Italian.

See https://support.mozilla.org/fr/search?q=l%E2%80%99accueil or https://support.mozilla.org/fr/search?q=l%27accueil
To partially worked around that, accueil has been set as a keyword in https://support.mozilla.org/fr/kb/comment-definir-page-accueil and https://support.mozilla.org/fr/kb/page-accueil-defaut-firefox-a-acces-rapide-vers-fonctionnalites-courantes for a search like https://support.mozilla.org/fr/search?q=accueil

Comment 1

[Kadir Topal [:atopal]]

Not sure if there is anything we can do here. Mike?

Comment 2

[Michael Cooper [:mythmon]]

ElasticSearch provides two French capable analyzers. One of them is Snowball, which has a setting for french, and the other is the analyzer simply called French. I assumed that Snowball would be the better of the two.

I did some quick checks, and French seems to deal with word combinations like d'accueil and l'accueil better than Snowball. I don't just want to jump at the first sign of trouble though, so I want to do some more research here.

However, modifying the analyzers beyond what they do natively would probably be more trouble than it is worth.

Comment 3

[Michael Cooper [:mythmon]]

Some more thought on this: We should get a sample of French search queries, and see how each of the analyzers deals with them. Then someone who speaks French can pick which one looks better. We should probably also do the same with some text of popular SUMO KB articles in French.

Ibai: Can you get a list of common searches in French on SUMO? I don't know how to work the Google Analytics interface well enough to find the information I want.

Comment 4

[Ibai Garcia [:ibai]]

I can:

https://docs.google.com/spreadsheet/ccc?key=0AiYrsx83iOr0dDdtejVXR2x3YjIxd0ZQOXFsRGNwWEE&usp=sharing

It's a pretty rough list. I hope it's enough. Let me know if it doesn't and I can get an unsampled report.

Comment 5

[Scoobidiver (away)]

I think there's a module 45 somewhere and spaces at the end should be removed to aggregate the same terms.
That said, it helped me to tweak keywords for common search terms and typo.

Does French contain common typo such as words without accent?

I will attach a file based on the one from comment 4 and that displays the assumed expected article and the current search result rank.

Comment 6

[Scoobidiver (away)]

Attachment: Expected results for French searches

Comment 7

[Seburo]

Closed as per https://blog.mozilla.org/sumo/2019/06/07/sumo-platform-roadmap/