Google Co-op

From Meta, a Wikimedia project coordination wiki

• Wikimedia Data test account at www.google.com/coop/profile?user=016597473608235241540

Google Co-op is a Google project which currently consists of two modules:

1. Topics
2. Subscribed Links

Both modules are interesting for Wikipedia content. This page is documenting any effort to prepare Wikipedia content for a Google Co-op.

A non-Wikipedia specific description of google co-op can be found at the wikipedia article "Google Co-op"

Contents 1 Subscribed Links 1.1 Having fun with Persondata / Personendaten 2 Having fun with Geodaten / Geodata 2.1 Distance calculation 2.2 Information about cities 3 ToDo 4 Credits 5 Bug reports 6 Code

[edit] Subscribed Links

Subscribed Links basically allow third parties to provide content in a more or less structured way to be shown to people above the standard results. The data is stored in a XML file.

[edit] Having fun with Persondata / Personendaten

Personendaten / Persondata are a method of storing information about biographic entries within the article in a template format:

{{Persondata
|NAME=
|ALTERNATIVE NAMES=
|SHORT DESCRIPTION=
|DATE OF BIRTH=
|PLACE OF BIRTH=
|DATE OF DEATH=
|PLACE OF DEATH=
}}

German Wikipedia introduced Personendaten in 2004. We used this data when we cooperated with Directmedia Publishing GmbH in producing the German Wikipedia DVD.

As of june, there are about 80.000 entries with Personendaten which can be downloaded at the Wikimedia Toolserver.

This kind of data is very useful in playing with Google Co-op/Subscribed Links:

<Results>

<AuthorInfo description="Gib das Geburtsdatum einer Person aus" author="Mathias Schindler"/>

<ResultSpec id="GeburtsdatumMatch1">
	<Query>Geburtsdatum [Geburtsdatum]</Query>
	<Response>
		<Output name="title">Geburtsdatum von [0.fullname]</Output>
		<Output name="more_url">de.wikipedia.org/wiki/[0.fullname]</Output>
		<Output name="text1">Das Geburtsdatum von [0.fullname] ist</Output>
		<Output name="text2">[0.dob] (laut Wikipedia).</Output>
	</Response>
</ResultSpec>

<DataObject id="5" type="Geburtsdatum">
	<QueryName value="Ang Lee" /> 
	<Attribute name="fullname" value="Ang Lee"/>
	<Attribute name="dob" value="23. Oktober 1954"/>
</DataObject>

</Results>

This is a very basic feature using the Query Pattern feature.

If you search for "Geburtsdatum Ang Lee" (or any other of the 80.000 names), you will get the following result:

(The url is broken and some aspects could be better)

[edit] Having fun with Geodaten / Geodata

[edit] Distance calculation

Thanks to Eric, there is now a great application to Wikipedia meta data: Calculating the distance between two cities:

[edit] Information about cities

Thanks to Stefan Kühn, I got a list of extracted information from geotags. This is a very simple case, everone here could think of more:

File:Google-coop-mannheim.PNG

<Results>

<AuthorInfo description="Find country of a given city" author="Mathias Schindler"/>

<ResultSpec id="CountryMatch1">
	<Query>Where is [stadt]</Query>
	<Response>
		<Output name="title">[0.fullname] is a city in [0.staat]</Output>
		<Output name="more_url">en.wikipedia.org/wiki/[0.fullname]</Output>
		<Output name="text1">[0.fullname] is a city in [0.staat]</Output>
		<Output name="text2">Wikipedia contains an article about [0.fullname] and [0.staat]</Output>
	</Response>
</ResultSpec>


<DataObject id="town2" type="stadt">
	<QueryName value="Ankara"/>
	<Attribute name="staat" value="Turkey"/>
	<Attribute name="fullname" value="Ankara"/>
</DataObject> 
<DataObject id="town3" type="stadt">
	<QueryName value="Aachen"/>
	<Attribute name="staat" value="Germany"/>
	<Attribute name="fullname" value="Aachen"/>
</DataObject> 

...

[edit] ToDo

Persondata

• How old is $Name? (current year minus date of birth)
• Where was $Name born? (place of birth)
• What is $Name famous for? (short description)
• Where did $Name die? (place of departure)
• When did $Name die? (date of departure)

Geodata

• Where is $Place?

Plants

• What Genus does $plant belong to?

Towns

• How many citizens does $town have?

[edit] Credits

I have to thank Jakob Voss for providing Persondata files, Stefan Kühn für Geodata files, Magnus Manske for preparing leadsections of a bunch of articles. And many more people. Thanks for the Googlers for creating this lovely tool. -- Mathias Schindler 21:18, 11 June 2006 (UTC) And JeLuF for high-speed-coding. -- Mathias Schindler 12:27, 18 June 2006 (UTC)

[edit] Bug reports

• Bug 6294: Support for the most basic structure of Google Co-op

[edit] Code

• Mediawiki SVN:ActiveAbstract/GoogleCoopFilter.php