How do we ask for what we don’t know?

It’s my favorite question. I chew on it all the time like it’s a mental cinnamon toothpick because it’s such an essential question of search.

All we can bring to a search engine topic query is ourselves and what we know. And unless we’re some kind of expert on the topic, we probably know woefully little and what we know probably contains misinformation of various sorts. Our lack of understanding makes us more susceptible to junk searches or shallow, SEO-serving search results that exist only to bump up a Web site’s ranking, and not to provide knowledge.

Knowing everything is an impossible strategy. Fact-checking every single Web page you get in a search result, also impossible. So currently we tend to trust Google (or Bing or DuckDuckGo or You or whichever search engine you use) to guide us to the most useful search results available.

The problem with that is twofold: search engine algorithms are usually opaque and there is a constant conflict between the search engine trying to serve the most useful results possible and SEO black hats trying to game the system and serve results for the money/propaganda benefits/etc.

I can’t start a search engine because ResearchBuzz is just one (1) person. But I can and do try to figure out ways to make searches focused enough to break through the SEO / general knowledge fog and into richer results  with more context.

In December I made Clumpy Bounce Topic Search, which was an attempt to use Wikipedia categories to build Google queries for broad topics. And it works pretty good and it’s fun, but it’s no good for specific topics, people, etc.

I’ve been trying a bunch of different approaches to address this, to build a set of related query topics that works and provides meaningful search results without excessive weirdness and junk results. Finally I figured out a good way was to count Wikidata entry mentions across Wikipedia pages, so now there’s Wiki-Guided Google Search ( https://searchgizmos.com/wggs/ ).

Using Wiki-Guided Google Search

To use WGGS, you provide two things: a Wikipedia topic (it’s case-sensitive, so keep that in mind) and the number of times the topic should be mentioned in another Wikipedia article before that topic is included in the search results. Fewer mentions will lead to less-associated topics (and occasional nonsense.) If you’re not sure how many mentions you should screen for, start at 2 and go higher if you’re getting too many results.

Let’s stick with the default search here. Solar energy is definitely a popular topic, so it’s going to have lots of mentions. A mention filter of 5 will still find plenty of results. Click the search button.

Results include the name of the associated topic (with a link back to its Wikipedia article), a bit of excerpt, and links to Google and Google News searches for both your original topic and the associated topic.

The first time I ran this search I said out loud “Morocco?!” Morocco as a topic associated with solar energy would have taken me a while to come up with on my own, though it makes sense if you think about it. And man, does it bring great results.

Adding Morocco as additional context to our query about solar energy gets past those shallow sites about solar power and takes us straight to rich results (and, of course, a Wikipedia article.) Here’s another result, this time for “solar energy” and “Passive solar building design”:

I don’t have any proof to back this up, but I suspect that the formality and structure of Wikipedia’s language use helps make the results more information-oriented.

WGGS also works great for people search, especially for people who have been influential in our culture but are possibly lesser-known. My favorite pianist is Henry Byrd, who used the name Professor Longhair. He influenced any number of better-known musicians but is not well-known himself. However, if you put his name into WGGS with a filter of 4 you’ll see that his name appears in a variety of contexts:

And these too yield tasty search results.