Last modified: 2014-05-31 16:00:27 UTC
I recently searched for media (images) about "master race" on commons. Master race is a concept going back to the 19th century, that "one race" is superior to the others. Instead of what I wanted, I was given various images of "races", so race-cars, drivers of race cars, bikes, ... I therefore propose the following "search algorithm": -First search for the phrase as it was entered -Then search for modifications of the phrase (eg. all lowercase, ...) -Then seearch for different parts of the phrase. For each hit you get, you should keep track of which of the three cases it originated from. The UI can then be modified, to let the user "toggle" the different kind of hits he gets. In addition, very basic stats could be displayed (as in: 15 hits for case 1, 22 hits for case 2, and 200 hits for case 3).
Please add some measurements about the suspected performance impact in comparison to the current system. (This will probably need testing on a larger scale)
I go to https://commons.wikimedia.org/ I enter master race in the search field and hit the Enter key. Results are as described by Robert. If I enter "master race" (note the quotation marks) I get 47 results, the first 45 about "masters race" showing fast cars or drivers of such, and the last 2 refering to "the racialist theory of the Aryan master race" of the "Arthur de Gobineau" article. Entering "master race" -masters (the minus in front of a term implies to ignore this term) in the search field I only get 5 results and NO results about "masters race". So the latter is likely what you want and the workaround, but it's not easy to find out about the options to use quotation marks and minuses for search terms.
I'm inclined to WONTFIX this as it's all working as intended. Basically you're wanting exact-string matching which is what quoting it does. Changing the behavior would confuse a lot of people. searching for `master race` means you're searching for `master` AND `race` Searching for `"master race"` does better, but you'll still get the stemmed results for `masters race` like Andre pointed out.
Makes sense
