-------- Wikipedia

This is the Wikipedia page of Ronaldo, the famous Brazilian Footballer.

The YAGO system extracts structured information from this page, and stores it in a knowledge base.

-------- YAGO

This is what the knowledge base knows about Ronaldo. 

It knows for example his birth place, his awards and his gender.

In our demo, we show visitors how the YAGO system creates this knowledge base.

-------- GUI

This is the entire architecture of the YAGO system.

The blue circles are extractors. They extract information from different sources.
The purple boxes are themes, that is collections of facts.

Let us look at one particular extractor, the Category extractor.

-- Zoom in on Category extractor

The Category extractor extracts information from the categories of Wikipedia,
in our example the Wikipedia page of Ronaldo.
-- show wikipedia source

Now we run the extractor,...
-- Run Category extractor

which produces the following facts:
-- Show extracted facts

These facts are not yet clean. For example, there is no Wikipedia page "European footballer of the year". The page is called "Ballon d'Or" on Wikipedia, and hence in YAGO.

This is why the facts have to be processed further.

-- Show dependencies of the extracted facts theme

The next extractor in line is the redirector.
It will take care to redirect the faulty name.

-- Run Redirector

-- Show resulting facts

 Now, the faulty entity name has been corrected.

-- Show dependencies of resulting facts

The facts are further filtered and merged, until they land in the final YAGO.
-- Show fact in the final theme
We can also show the provenance of the fact
-- show provenance
[...]

The user can also edit the Wikipedia file, and see how the system deals with the changes.
-- edit Wikipedia source

Here, we add the category "Cities in Italy".
-- type it, save, and close.

If we now run the extractor again, it will think that Ronaldo is a city located in Italy.
-- show output
Now we run the Redirector.
-- show dependencies of output theme
-- run Redirector
-- show output
(Oh, no!) The Redirector has not filtered this fact out.
But the next extractor is the type checker. It verifies the domains and ranges of the facts.
-- run Type checker
-- show output
As we can see, the typechecker has filtered out the erroneous fact.


The user can also add new extraction rules, new relations, or even new deduction rules.

Our interface allows the user play with the YAGO system and see how it works.
We hope that this can give us feedback from the visitors,
and inspire other researchers in knowledge base construction.