0

I am talking about the text that is shown right of search results on Google.com:

Google Search: search for "germany" shows information about Germany in the right sidebar

For institutions, universities this is often the first sentence from Wikipedia. For countries it is not.

When I search for the whole sentence

Germany is a Western European country with a terrain of vast forests, rivers and mountain ranges, and 2 millennia of history

there are only four places on the web that contain it, non of which is Wikipedia or any known institution.

Are these text snippets statistically generated summaries? If so, what information does Google provide about generating them?

3
+100

Here you can find a good explanation about the Knowledge Graph, the system used by Google to give "answers to search queries", and not "links to search keywords".

http://www.seobythesea.com/2013/05/google-knowledge-graph-results/

Getting the Best Summaries – by organizing and presenting information about an entity that appears within our queries in a meaningful way by studying in aggregate what people have been asking Google about each entity to be displayed in a knowledge panel. [...] How does Google know what to show about these things? One hint from the patent, echoed in the video, is that what people ask about in Google can determine what Google will display in the knowledge panel. [...] Knowledge Panels may contain information from a social networking page related to an entity as well as the other types of content I’ve mentioned above.

http://www.lunametrics.com/blog/2012/05/22/making-sense-googles-knowledge-graph/

Where does the data come from? According to Google, the Knowledge Graph is derived from an enormous pool of diverse data sources. Google literature has specifically made mention of the following: 1) Wikipedia 2) Freebase – incidentally acquired by Google nearly 2 years ago, and also the only source in this list linked to in Google’s own literature about the Knowledge Graph. 3) Subject-specific sources such as Weather Underground, World Bank, and CIA World Factbook. 4) Google’s own immense stores of search data.

So, it could use lot of different sources, and also combine them. It evaluates not just the domain authority of content on a webpage, but even the social signals among them. At a glance, that description seems to be taken from some paper book or touristic website: it's a romantic and promotional way to describe a country, not an encyclopedian one. But maybe I'm wrong and it's just an old Wikipedia entry. What we know is how Knowledge Graph works in general, and the phrase you blockquoted is in more than 100k results in Google's index.

This one

Berlin, its capital, is home to thriving art and nightlife scenes

With "" quotes (phrase match) is in only 5 results. But if you remove the quotes, you get 600k results. So, there are tons of pages where Google takes data.

By the way: if I search countries on my tld, google.it, I get every time Wikipedia entries.

| improve this answer | |
  • Thanks for the answer. It's clear when the text snippet comes from Wikipedia, it's marked as such. No problem. What puzzles me is how exactly is the snippet generated when the source is not Wikipedia. Yes, when you remove the quotes there are thousands of results but these are a lot less credible than Wikipedia or some other Encyclopedia, well I would not expect Google to put quantity before quality but still, question remains, can this snippet be influenced in a similar way that page ranking does? If so, how? – matcheek May 20 '15 at 14:39
  • lunametrics.com/blog/2012/05/22/… I'm adding it to the answer. We cannot know how Google exactly works. The only thing we know is thanks to testing, Google's own words and their patents. Nothing else. For Google, quality is also condensing hundred of variables and giving results based on them, not just give answers based on the same sources, even authoritative ones like Wikipedia. – lucgenti May 20 '15 at 15:09
  • I took a quick dive into Freebase API and after checking a few countries I can tell the the results differ a lot from what gets presented on Google.com. Knowledge graph is a vast topics but I am only interested in part of it. There are only under 200 countries, so strictly speaking an office clerk would write all of them in a week. Yes, there are plenty of signals but search rankings mechanisms are either made public directly by Google or figured out later by the SEO businesses through testing, so I believe that I can get to the bottom of it. – matcheek May 20 '15 at 16:27
  • 1
    They can made public if results are obvious. If they show you Wikipedia, they can tell you "Google found Wikipedia trustworthy and so Google show you their description". They have interest in explaining main functionalities, but without the details, as always Matt Cutts did in the past. I've never found more about the topic, and as you can see the sources posted are valid, in particular Bill Slawski. Look at that also seobythesea.com/2014/10/… If you find something in next days, let me and rest of the community know. – lucgenti May 20 '15 at 17:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.