Why does wikipedia's search *really* suck?

I’ve wondered this for a while. It’s downright atrocious. What gives?

This seems to come and go in waves, perhaps they are optimizing their search function coding, and making it worse for a little while. The last time it happened, I was sure they’d nerfed it deliberately to save some server load. But then they fixed it. Meantime, you can get an excellent Wikipedia search by using Google, just tack “wiki” after your search term. You might even discover another, specialized wiki, that may be better then Wikipedia.

Or, if you want to search english wikipedia specificially, you can add


site:en.wikipedia.org

to the search terms. Also works for any other site if you’re interested.

Search is difficult and expensive to scale up. Google has set the bar very high.

It’s not that hard to make a mediocre, slow search engine. But it would take more resources than Wikipedia has to build one that’s good. There’s also a feedback loop: the better it is, the more people will use it. Which means it consumes more resources, and thus needs to be better still.

Wikipedia is non-profit and, as far as I’m aware, generates no revenue apart from through donations. A few years ago it only had 5 employees - I think it’s still fewer than 50. And I can’t imagine they get paid very much. They just don’t have the resources to create a search function closer to the level of Google, which has thousands of employees and makes billions of dollars in profit each year. It’s astonishing that Wikipedia is as usable as it is to be honest.

Depending on which figures you look at, Wikipedia apparently gets something like 30,000 searches per second. They’re probably not trying to avoid getting “too many” hits!

Or maybe just do it without the “wiki” part and you may get the original source rather than the paraphrased or outright plagiarized wikipedia version. That’s one of my pet peeves—“citing” wikipedia. Wikipedia isn’t a cite, it is a list of cites.

Hey, at least they now sort of have a spell checker. Used to be if you got one letter wrong… nada.

Wikipedia’s fine for what it’s meant to do. Don’t cite it for your dissertation, sure … but if you’re interested in XYZ, the Wikipedia article on XYZ is hard to beat as an initial stop.

And, as a list of citations, it is inherently superior, since you get more than one view.

One of my pet peeves is people thinking they are superior because they dislike Wikipedia. It’s like the old style of treating Internet research as if it isn’t research. All research is dependent on the skill of the researcher, not on the way the information is presented.

And a Google search gives you more than one view as well, plus you aren’t tied to a list of what one person (or at best a few people) decided to list.

Typically, you get a list of websites that spent the most money on marketing to fool secret algorithms programmed by “a few” engineers. That, or you get sent to the Wikipedia article itself.

Or you can use Google Scholar and get a list of half a million semi-relevant articles and spend the next three weeks reading through them all to get an idea of what’s most important.

Or you can reuse the work of “a few” dedicated, interested volunteers who did all that for you, hand-selecting sources and outlining and summarizing them for you, and THEN go to the primary sources, finding more later if that initial batch wasn’t enough.

But all of that is irrelevant, since the point of this was the search engine not being good in the first place.

Very true. I wonder why we’re talking about citing Wikipedia in the first place. Hmm…

Because I made an aside in a post about using Google instead of wikipedia as a search engine, and the thread then got hijacked.

Ok. Back on topic then.

Some of the reasons Wikipedia’s search isn’t all that great have already been mentioned.

But there’s one way we can all help make it better for everyone: Every time you type in a relevant article name and find nothing (only to find the right article later on by browsing or by Googling Wikipedia), take five seconds to make a redirect of your original query to the right article. Anybody who searches for that phrase in the future will get pointed to the right place.