36C3 Wikipaka WG: Live querying: let’s explore Wikidata together!

Music

Herald: We'll do some live querying with him
so you were told to think of some ideas that

we could search for in Wikidata and when
we get to that point I would ask you to

raise your hand and wait till I get to you
with the microphone so that the people in

the stream can also hear what we're
talking about so that's the thing I'll go

back to Lucas and we still have
translations. Wenn ihr es auf deutsch hören wollt

wir haben immen noch Übersetzter, die alles versuchen
es Euch nochmal auf Deutsch zu erzählen. Also hört mal rein.

Lucas: Does anyone have a query?
Yes, in the front there. We have a question allready.

Audience member: Is it possible to find all circular family trees?

L: All circular what?

AM: Family trees
L: It's certainly possible to find

some. Finding all it's probably going to
be a timeout but there would be something

like select, probably child would be
the simplest, so item child plus item

again. So if we put the star like earlier
then you could then every Tree would

match that, but with the plus it means it
has to be at least one child link or more,

and let's just add a "limit 1" because I'm
not that optimistic that this is even

going to find one, but I'm pretty sure we
cannot find all of them, but let's see if

we can find one and this might just take a
while, but I don't think there is a good

way to do this otherwise unless you
download one of the dumps either the JSON

data dumps or the RDF dumps which is the
same data format used here and then you

can do it locally without an timeout.
I don't think there's much I can optimize

about this query is pretty short unless
like I had an idea that people named John

are more likely to have these kinds of
cycles, then we could filter it down first

but you men. I'm afraid that is not going
to work it looks like. Yes, timeout. And

you can see the thing is written in Java
the server dragazines. One thing we

can do with this "P40+" is something like
search start with a certain mythical

creature such as King Arthur. I hope I can
find him like this. Search is being alright

3d map of EDT. There we go, that's the
legendary British or Welsh king and then

we are searching for an item who is
definitely a real human and who has a date

of birth and we say the date of birth
should be greater than at say 1950 and

this is a date time value and this let's
even say 1980. I think that might be

No more efficient. There we go. No
results, okay. I thought King Arthur had

some real descendants. Though then it was
some other mythical creatures. Let's just

start with any ancestor who has the item
as child and the ancestor is also instance

of mythical creature, mythical character.
Let's see if we have any mythical

characters with children who are born
after 1950. Oh I still have the "limit 1"

here could make that a "limit 10" probably
or something, but I'm optimistic I think

there are some people here, especially, I
think, even British MPs, there's some

that's already on the list of example
queries British MPs with mythical

ancestors and there's lovely have traced
their lineage back to some 6th or 5th

century person and you have all the
apparent links in there and then it's kind

of tricky to figure out where it starts
being wrong. Oh that's not working out so

well. Does anyone else have ideas in the
meantime? There, way in the back.

Someone: Thank you
Audience member: We all know that stupid

game in Wikipedia where you try to find
the Adolf Hitler page by only clicking

links, so can you find the number of pages
that are directly connected to the Adolf

Hitler page in Wikipedia?
L: You can. Oh that was a timeout, dammit.

So that would be kind of ... It's a one funny
story about that for example is there's

the first example we
have here is cats and why do we have cats

and not dogs? Because if you search for
dogs the second result, no, it's the

fourth result by now, but that's the
thought of Hitler and we don't really want

that and normally so we usually use cats
as the example instead but let's just

search for anything where the item has any
connection and we don't care which

property it is to Adolf Hitler, like that,
and we are going to find 920 results. ok

some of these are site links so we also
want the item to have some label which

uses this new namespace and we want only
the English label so the language of the

label should be in English and we then we
just select the item and the label and

hopefully that's still pretty efficient.
Here we go NSDAP membership number

that's actually a property but I assume it
has as the example yep there's a property

example here as NSDAP number 1. World War
two has probably of cause of death do we

have him as an example on cause of death
really? and we have nitric acid poisoning,

stroke, cholera, shot to the head, cyanide
poisoning, hanging, That's a very pleasant

list. Do we need to have that many
projects handfuls of closet yeah then we

have Nazi Party, Klara Hitler, I don't
know who that is, 1936 Summer Olympics,

all kinds of things, so that's how
you can find all the things with a direct

connection to Hitler. Any other
examples? Yes, over there in the right, or

was there already somewhere someone back
there that I missed

AM: Can you find the
cheapest public infrastructure projects in

Germany?
L: The cheapest public infrastructure

what?
AM: Projects

L: Projects
AM: Like a bridge building

L: I don't think we're going to have a
full dataset about that but you can try.

Let's start with a more expensive one and
[crackling noise]

L: see - perhaps move away from the box,
that might help. Let's start with a very

expensive project and see just what the
data model looks like so what does

infrastructure project look like what's
what was the cost so the cost is probably

going to be in euro and I don't know how
to write here over there okay it's a

property called cost in Euro and does it
have something like instance of

international airport, building under
construction, Greenfield Airport, proposed

airport being built, so we could check
first is Berlin Brandenburg Airport, is

that an instance of some subclass of
public infrastructure? Is that a thing?

that looks like the wrong item what is
this this is nothing. Okay. There's anything

linked to this item? no nothing like
suicide. Okay. So it could be an

international airport is a subclass of
airport which is a subclass of an

aerodrome which is an architectural
structure, and we can search for

architectural structures, so the structure
would be an instance of subclass of

architectural structure, and it would have
a cost, and order by the singing costs I

think it's 10 and we're probably going to
get things in like yen or some other

currency where this number is just going
to be very high because we're not taking

any conversions into account right now but
let's see if we find something there. What

is it doing? Okay... not sure why this is
taking so long. Let's try a second version

in the mean time where we quantity amount
is cost and various quantities units

should be the euro they're still running
and yeah let's try this that works any

better or not? Okay, this was a timeout. This
looks like it's going to be a timeout as

well. I don't know, we can just search for
the most expensive things at all. Remove

this part, there we go. This costs 55
billion euros. What is this thing? Power

of Siberia, natural gas pipeline. That's,
that's in euro, the costs? Apparently. And

then this is 15 billion euros and then
8.77 find something that's the channel, oh

the Channel Tunnel is expensive. The
Brenner Tunnel was also expensive.

laughter
And Stuttgart 21 took about 21 whatever

was also- or is projected to be expensive.
Do we have one cost or several? Okay in

2018 we have a cost of 7 billion. Yeah, so
let's sort by the ascending constant set

because that was what we actually wanted
and then we get... okay now we're going to

get a lot of things that aren't really
infrastructure projects we have the whole

and a hot and energetic universe. Does
that mean it's a no budget film or what?

Okay. So we would need some kind of ...
Let's say, let's do duck typing instead of

saying it is an infrastructure project,
let's say it has, I don't know, a

coordinate location. And if it has a
coordinate location, we're going to call

it some kind of infrastructure project, or
at least it's not going to be a

documentary film. Perhaps that works
better. Yeah, so 21,000 euros cost this

thing which was in France. Oh, okay,
right, it should also be country Germany.

Here we go. That's 400,000 euros for
fountain in Stuttgart. Does that count? I

guess. And that's the engines of something
it doesn't even have a German la- an

English label, just a German one. Wait...
Oh, so this is the class of all the

fountains with exactly this name which are
a subclass of well and are all named after

this goddess, okay, cool. Yeah so then we
have some of these cheap projects, which

is… this public square… a bridge – oh
yeah, there's this tiny bridge, a

footbridge, has even an image, that's what
it looks like, and it costs, what was it,

1.6 million euros already. Wow. And then
we have another public square. Yeah. So,

"cheap public infrastructure projects".
And also probably "infrastructure" in

quotes, because we're really just saying
it has a location and "Country: Germany".

And, yeah, I can send this query around
afterwards. And this didn't work, this

didn't work. Okay, any other ideas? That's
bad news. We could try to continue with

some of these. Was there something? Oh,
from the Camera Angel!

AM: I have a question! I saw that with
Wikidata Query Service we can draw these

nice trees and have images in them, and
one example that came to my mind was all

the programming patterns – programming
design patterns, but grouped by their

kind, like they're structural patterns,
convenience patterns, and so on, and like,

can we draw a graph and maybe put an image
in them.

L: We can try that. So let's see how
that's modeled, I don't know, with the

visitor pattern for example. That's a
design pattern what kind of statements

does have. It's a subclass of behavioral
pattern, is this a programming thing or

already…? Oh yeah it's a soft… okay it's a
software design pattern. So we should say ...

We're going to have a pattern with
its label and a pattern kind with its

label and the pattern is going to be a
subclass of the pattern kinds, which is

going to be some subclass of – what was
it? Of software design pattern – and I'm

just going to copy this ID so it's the
right one – label service, and say, I

would like to see this by default in the
graph view. Here we go. Well that looks

not as bad as I thought. We have a lot of
structural patterns, behavioral patterns,

one architectural pattern, a few
creational patterns, and one fundamental

pattern. Yeah. And… yeah what we could
also do is, if we do this, then we should

also see connections of all of these.
Now we have the tree rooted at

software design patterns, we have monads,
and fundamental pattern is a kind of

software design pattern. Structural
pattern… and it's all linked there and

this is working… very well, I… That's much
better than I expected. I expected a huge

mess of… because it sometimes gets
different to determine when should you use

"instance of" and when should use
"subclass off", like if it's software or

patterns like this, I expected we would
have to account for both of these, but

this looks very good to me. I think we
don't need to do anything with this query.

Yeah, so that is, uhm, software design
patterns by a pattern tree.

Okay. Any other ideas? Or I can try to
keep optimizing this one

AM: Which cities have applied to be host
city of the Eurovision Song Contest the

most times but were never successful?
L: Oh!

Laughter from Audience
L: That's a very good question. I don't

know if we have– do you know who applied
for this year or for some year? But I

could check if the state if that's modeled
anywhere. Uhm, I need some example cities

so… let's check ESC 2018 if it has
information on where it took place, which

one won the bid, but also who was
nominated or something, or who applied… We

have "presenters", we have "followed by ",
"start time", "end time", "participants",

we have the winner, do we have a location
at all? Oh yeah, there it is. Okay, we

have a country, and a location, but I'm
not seeing any other countries here, and I

assume that information is not going to be
on the country item. It's possible that we

have some separate item for "Eurovision
2018 Bid" or… Well wait, it would have to

be "which city", because the country is
determined by the winner isn't it? So the

city, but I suspect we don't have that
information. We have a list of host

cities, but that's just… a Wikipedia list
article.

Interference noise
Do we have to switch to the other mic? Oh

no, that sounds great! Okay. Yeah, so we
don't have any of the structured

information here. It's just linking all of
these Wikipedia articles together, and

then here is the actual list with the
different venues. But I don't think we

have that information in Wikidata at the
moment. We could add it, you'd have to

figure out the data model, but it would
probably be relevant enough, I think.

I wonder if we have that for the Olympic
Games. So, Olympics 2020, do we have the

process of who applied to host those? Uhm.
We have a location. We have parts. Let's

check. Perhaps English Wikipedia has a
separate article about the selection

process for the 2020 Summer Olympics.
Doesn't look like it. "Host city

selection". No I don't see a main… oh no,
there! "Bids for the 2020 Summer

Olympics", that's the Wikipedia article.
Does that have any useful information on

Wikidata? Bids for Olympic Games no.
Damnit. So you can see, when these bids

all happened, but we don't have the
bidding countries and cities apparently on

Wikidata, at least not as far as I can
see. Bids for the… 2012 for example…

No, sadly, we don't have that information
yet. Did this one run by way? No.

Any other questions?
Herald: Our translation angels had a question.

H: They want to know, if can give them the
countries with the most colorful flags

L: Yes! That [interference noise] should
be possible. So "select country", and the

"count of the colors as counts"
[interference noise] the country has,

oops, has a flag, not the "flag image", a
flag, and the flag has color. And it

should be "color" and not "colors", and
then we group by country so this is a bit

like a [noise] grouping and aggregate
functions

[Interference noise]
Interferene noise*

Do we need to use the other microphone?
[Noise] Okay [Noise] But then you can't

really walk around anymore.
H: Hello hello? Hello hello? Muss man da

noch was machen?
L: Okay so now… This could be really fun!

Yeah, so we are searching for countries
with flags, and hope that the flags have

colors and been counting them, and what I
didn't do is… what's this? Do I want to

know? Okay, okay it's at least it's not
the straight pride flag, I guess. Does

this have 14 colors? No, what was it? No,
eight, I guess, one, two, three, four,

five, six, seven, eight, yeah. That's
accurate. Yeah, I didn't filter for

countries here, the thing is, country is
really a stupidly complicated term, so

what I did was… queries… I have a pre-
prepared query for the UN member states

somewhere, which I just copy all the time.
And this is now going to be called a

state, and then we only get state flags,
uhm, and there's exactly– oh, right. I

need to group by "state" and "state label"
and copy these up here as well, and then

it will hopefully work, and we will find
out that… the United Kingdom has… 12?

I suspect that's because it has four flags,
which all have the same rank, or a no– no

it should be five, right? United Kingdom
and Northern Ireland, Scotland, Wales and

England. Let's search for "flag". Flag is
the flag of the United Kingdom, no? Why

does it have 12 colors? It has blue, red,
white… wat. I see. But that still doesn't

explain the 12. Let's count only the
distinct colours "distinct", there's auto-

completion, thank God, perhaps that helps
. Though I don't know why it would have…

oh it would have had the state multiple
times because it's a sovereign state

multiple times probably. Let's check. Yeah
the United Kingdom is, it's a Commonwealth

realm, and an island nation, and a
sovereign state, and that's probably why

we got it multiple times, and, yeah that
looks more reasonable. South Africa,

Ecuador, South Sudan, and what we can also
do is, add the, of the flag, the image and

call that I, because I can't be bothered
to type the whole thing, and add that

here, and also add it to the "group by",
because otherwise it's not the right

aggregate and I can't be bothered to write
"sample" with one hand, and then we can

hopefully also see it. Oh, we get two
images of the flag of South Africa. That

also looks like one of them should be
"preferred rank", but anyways, we can

switch to image grid, and then we can see
all these colorful flags. One, two, three,

four, five, six, yeah. That's six. And
this is more than six, so I guess, I would

say that should actually be two separate
items, for this old flag and– no, this old

flag and the new flag, but… This is six…
is that only six colors? I'll believe it.

This is six colors, six, and then we have
five colors, yeah. So here are the, let's

just add a comment there, and I will tweet
this out later as well, "colorful state

flags". Yeah. And also we can use
the image grid as the default view.

We probably have time for one more question,
if it's a short one. Though I won't be

able to type very fast. Yes, let's
hope this works. Otherwise I can repeat it

for the stream if I hear you.
AM: So does it work? Yep seems so. I don't

know if it's possible, but the smallest
images that are on Wikipedia? So, by image

size?
L: That would not be possible with the

Query Service I think. But I think on
Commons you can search… can you search?

Whoops, I don't have that search shortcut
set up here. Can you search by image size?

I think that might be possible. Advanced
search, file type, sorting order… No.

You could probably sort by a file size in an
SQL query. Which is not a thing from the

Wikidata Query Service, but it's possible
with something else and as it happens I am

going to have another talk later, where I
talk, about among other things, how you

can write SQL queries against the
Wikipedia databases, and then we might be

able to find a solution for that, and
that's I think at 6 p.m. today over in the

Esszimmer, or you come over to me after
the talk and then I can try to figure it out there.

H: A last emergency idea that we have to
try out?

H: I'm muted. Do you have ano– one more
idea? A small idea maybe we could do but

other than that I think we are, so– filled
the time quite well.

L: Yeah I think we're done. But if you
have any other ideas, you can always

contact me on Twitter @wikidatafacts, or
on Mastodon as well, and then I will see

what I can do for you. Yeah. Thanks.
H: Thank you very much, Lucas, that was a

great introduction to Wikidata querying!

Music

Subtitles created by c3subtitles.de
in the year 2021. Join, and help us!