As a non-English user, I want the quality of search completion to not be degraded by new completion on Commons.

The new Query Completion project (see parent task) will introduce a new Query Completion mechanism, based on past search interactions. Since most of the data (most of the searches on Commons) is in English, there is a risk that the quality of completion will decrease for non-English languages. This is mitigated by falling back on the default mechanism when the new mechanism does not provide a completion. This still needs to be validated by real data.

How to validate isn't entirely clear. We could sample a number of searches manually, or do some more in depth analysis (which might require a way to tag the language of a query).

AC:

• high confidence that search experience isn't deteriorated for non-English users after the introduction of the new Search Completion