Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Wikidata

wikidata.org

Post reply on HN
Print view View original

41–50 of 69 posts

Re: Wikidata

#41

Post by spupy »

mkr-hn wrote: post #15
leoxv wrote: post #13

Earlier quoted context omitted.

"Introducing the Knowledge Graph: things, not strings" https://blog.google/products/search/introducing-knowledge-gr...

I wish Google would just search for strings the 60% of the time it doesn't know the thing I'm looking for but really, firmly, and persistently thinks it does. Like it used to. It should get the hint and switch methods somewhere around or before the 10th page of useless, mostly identical results.

Isn't this kind of the purpose of the 'verbatim' search setting?

Re: Wikidata

#42

Post by zajio1am »

HelloNurse wrote: post #28

Tests are a bit depressing. - "Borane" has two unlinked entries: Q127611 is "any chemical compound composed of boron and hydrogen atoms only" while Q15634214 is specifically boron trihydride. Both correct, but the latter should be labelled as an instance of the former. - "Anomalocaris" has Q37395 for the "extinct genus of radiodon" (instance of "fossil taxon") but species (e.g. Q49557506 Anomalocaris cranbrookensis)…

> Both correct, but the latter should be labelled as an instance of the former.

No, it should be labelled as an subclass of the former. Instance would be a specific piece of borane.

Re: Wikidata

#43

Post by alyizzet »

Normally when you hunt for information that you can't absolutely find an already-made API of you scrape the web and deal with regular expressions but this is definitely the easier option, using SPARQL. I can see a lot of potential information harvesting use cases for it.

Re: Wikidata

#44

Post by ernesth »

chris_wot wrote: post #14

I have frequently created articles and had Wikidata linked to the article. Take https://en.wikipedia.org/wiki/Kate_Baker - if you click on Wikidata item, then it takes you to: https://m.wikidata.org/wiki/Q21536118

Nice.

It seems like magic, but it is possible to entirely outsource the infobox to wikidata. See for example https://fr.wikipedia.org/wiki/Mart%C3%ADn_Abadi whose infobox is created with the sole `{{Infobox Biographie2}}` line.

Re: Wikidata

#45

Post by M2Ys4U »

aasasd wrote: post #20

Alas infoboxes on Wikipedia often have more info. Particularly, in the case of cultural phenomena people are frequently eager to populate the infoboxes. Perhaps some kind of synchronization would be feasible, where with standardized infoboxes relations between known items could be extracted when specified on Wikipedia, and plopped into Wikidata. For myself, I looked around for a service that could present infobox-dat…

Wikipedia infoboxes can (and in small numbers are) being populated by Wikidata information. Some languages are making more use of this than others.

Re: Wikidata

#46

Post by SeanLuke »

It appears that Wikidata semantic network statements are untyped and binary rather than n-ary. Ugh. Amazingly, the above URL has a picture which shows exactly why this is such a stupid approach. The location of San Francisco(Q62) is a "Geolocation" consisting of a latitude and a longitude. That is, the statement is (if I'm reading this right):

  location(San Francisco (Q62), Geolocation{-122.4183, 37.775})
This requires that Geolocation be a special type like integer or float. VERY BAD. There would be a massive proliferation of such things.

Alternatively Wikidata could do this as

  Q95 (or whatever) = new object
  longitude(Q95, -122.4183)
  latitude(Q95, 37.775)
  location(San Francisco (Q62), Q95)
This is the standard reified approach used by binary semantic networks and is ugly as sin, unnecessarily polluting the object space with little object poops. The clean way to do this would have been to declare location to be ternary, so we could say:

  location(San Francisco (Q62), -122.4183, 37.775)

Re: Wikidata

#47

Post by ArubaJamaica »

leoxv wrote: post #10

Wikidata is a fantastic data source for all sorts of applications, not just for the Wikimedia projects (infoboxes, templates, etc.). I'm also excited about the Wikifunctions project ( https://en.wikipedia.org/wiki/Wikifunctions ) to created a large body of executable snippets that can be applied to data types. Wikidata is one of the main data sources in my Conzept encyclopedia project: https://conze.pt ( https://twit…

Like the UI a lot. Very interested to know what kind of user base you currently have and what use cases you've built it for.

Re: Wikidata

#48

Post by tshaddox »

mmarx wrote: post #5
uniqueuid wrote: post #2

Related, see this older attempt to create a structured dataset from Wikipedia rooting back to the semantic web: https://en.wikipedia.org/wiki/DBpedia "DBpedia (from "DB" for "database") is a project aiming to extract structured content from the information created in the Wikipedia project."

Unlike DBpedia, Wikidata is not structured content extracted by parsing Wikipedia articles, but instead a structured collection of factual knowledge created by human editors.

Human editors, and many many bots.

Re: Wikidata

#49

Post by aasasd »

M2Ys4U wrote: post #45
aasasd wrote: post #20

Alas infoboxes on Wikipedia often have more info. Particularly, in the case of cultural phenomena people are frequently eager to populate the infoboxes. Perhaps some kind of synchronization would be feasible, where with standardized infoboxes relations between known items could be extracted when specified on Wikipedia, and plopped into Wikidata. For myself, I looked around for a service that could present infobox-dat…

Wikipedia infoboxes can (and in small numbers are) being populated by Wikidata information. Some languages are making more use of this than others.

I'm talking about transfer in the other direction, or bidirectional one. Because people fill the infoboxes way more than they do Wikidata.

Re: Wikidata

#50

Post by mmarx »

SeanLuke wrote: post #46

It appears that Wikidata semantic network statements are untyped and binary rather than n-ary. Ugh. Amazingly, the above URL has a picture which shows exactly why this is such a stupid approach. The location of San Francisco(Q62) is a "Geolocation" consisting of a latitude and a longitude. That is, the statement is (if I'm reading this right): location(San Francisco (Q62), Geolocation{-122.4183, 37.775}) This require…

Wikidata statements are not binary, and values are not untyped. However, there is the concept of “simple statements”, especially in conjunction with the RDF exports (although not untyped, these data usually look like binary relations). Indeed, any statement can be annotated by a list of “qualifiers”, each of which can have zero or more values, can have references (which can have their own qualifiers)

As for types, “GlobeCoordinateValues”[0], i.e., locations, form one of the types allowed for values, and consist not only of two integers, but indeed of four distinct values: latitude, longitude, precision, and the reference globe, since Wikidata does not limit coordinates to the Earth. There is also no “massive proliferation”, the data model[1] knows 12 types[2], of which four are four different kinds of texts (untranslated, monolingual, multi-lingual, and list of multilingual texts).

[0] https://www.mediawiki.org/wiki/Wikibase/DataModel#Geographic...

[1] https://www.mediawiki.org/wiki/Wikibase/DataModel

[2] https://www.mediawiki.org/wiki/Wikibase/DataModel#Datatypes_...

Post reply on HN