Talk:Google Knowledge Graph
This is the talk page for discussing improvements to the Google Knowledge Graph article. This is not a forum for general discussion of the article's subject. |
Article policies
|
Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL |
![]() | This article is rated C-class on Wikipedia's content assessment scale. It is of interest to multiple WikiProjects. | ||||||||||||||||||||||||||||||
|
![]() | Daily pageviews of this article
A graph should have been displayed here but graphs are temporarily disabled. Until they are enabled again, visit the interactive graph at pageviews.wmcloud.org |
Title
editI have reverted a bold move to the title Google Knowledge Graph. The formal title of this feature seems to be just Knowledge Graph and this title seems clear and simple enough for our purposes. For an example of a mainstream source which follows this usage, see Washington Post. Andrew Davidson (talk) 09:03, 18 May 2012 (UTC)
- Random opinion here, it seems crazy to allow general purpose terms to be claimed by specific implementations/products. (I found another irritating case, 'bitplane' a company name, requiring annoying disambiguation when you want to link to info about certain bitmap formats. At the very least 'knowledge graph' should be a generic disambiguation page pointing to here, then in future people accidentally linking this will know Fmadd (talk) 17:22, 12 June 2016 (UTC)
- Agree, I would fully support a move back to Google Knowledge Graph. Crumpled Fire (talk) 17:40, 12 June 2016 (UTC)
- Barring another topic arising of that name needing an article space and thus disambiguation, it looks to me like this is the most appropriate place to keep it, as the most common term for it. --Nat Gertler (talk) 18:16, 12 June 2016 (UTC)
- IMO.. not even Google should be able to get away with claiming common descriptive terms as brand names IMO. there were plenty of tools for doing this sort of thing and it's a general purpose term, it just happens to be a technical term that wasn't previously in mainstream use.what about microsoft "Paint". you want every reference to 'Paint' to have to explicitely specify that it's not the microsoft tool. Apple's spreadsheet is called "Numbers". should we change the wikipedia article to be about apple's spreadsheet and rely on a disambiguation for that...and so on. other products have knowledge graph components, as will future products Fmadd (talk) 18:52, 12 June 2016 (UTC)
- Barring another topic arising of that name needing an article space and thus disambiguation, it looks to me like this is the most appropriate place to keep it, as the most common term for it. --Nat Gertler (talk) 18:16, 12 June 2016 (UTC)
- Agree, I would fully support a move back to Google Knowledge Graph. Crumpled Fire (talk) 17:40, 12 June 2016 (UTC)
- Random opinion here, it seems crazy to allow general purpose terms to be claimed by specific implementations/products. (I found another irritating case, 'bitplane' a company name, requiring annoying disambiguation when you want to link to info about certain bitmap formats. At the very least 'knowledge graph' should be a generic disambiguation page pointing to here, then in future people accidentally linking this will know Fmadd (talk) 17:22, 12 June 2016 (UTC)
- Agree title is confusing. Google Scholar reports 1130 papers on "Knowledge Graph" prior to 2011. [1] Michaelmalak (talk) 18:37, 12 June 2016 (UTC)
- OK, the general purpose term "Knowledge Graph" is pretty much a synonym for semantic network, which already has an article. I suggest the following: 1) Knowledge Graph redirect to that article, 2) This article be renamed to Google Knowledge Graph, and 3) Google Knowledge Graph be added as another example to the Semantic Network article. As an alternative -- much less preferred by me -- an "Other uses of" template can be added to this article that points to the Semantic Network article. The concept of semantic networks goes back to the 1950s, and the synonymous term Knowledge Graph goes back to at least the 1970s. An article titled Knowledge Graph devoted to a specific 2012 implementation by a specific company of a concept that has been researched and implemented continuously for 60 years -- again, 1130 articles in Google Scholar refer to non-Google uses of the term "Knowledge Graph" -- violates the MOS principle of article title precision. https://en.wikipedia.org/wiki/Wikipedia:Article_titles#Precision "Usually, titles should be precise enough to unambiguously define the topical scope of the article, but no more precise than that." Michaelmalak (talk) 21:56, 13 June 2016 (UTC)
- maybe a compromise would be to make 'Knowledge Graph' itself a disambiguation page (->semantic network or google knowledge graph) then that will clean up any existing erroneous links (and appease the opposing views). But I would personally be happy to see it as a straight redirect— Preceding unsigned comment added by Fmadd (talk • contribs)
- I would fully support a straight redirect to Semantic network with a hatnote saying — "Knowledge Graph" redirects here. For the Google product, see Google Knowledge Graph.. — Crumpled Fire • contribs • 22:27, 13 June 2016 (UTC)
- as I understand the wikipedia platform, it has a handy feature that warns you about links to disambiguation pages: this means during a transition, the community can fix up any links that become 'semantically' wrong after the change . Thats why a disambig page might be useful; if you put a note on 'semantic network'... you rely on humans to click through and find its wrong , manually. Leveraging the link structure between pages to do work helps IMO. Fmadd (talk) 22:34, 13 June 2016 (UTC)
- Although that may be the case, technicalities should not be relevant in deciding these matters. That being said, I'd support a disambig page if it was what others preferred. My primary choice is the redirect, secondary is the disambig, and lastly the status quo. — Crumpled Fire • contribs • 22:53, 13 June 2016 (UTC)
- IMO I describe practicalities for transition, rather than technicalities; the issue is the accuracy of the 'state' of information in wikipedia at any one time. Perhaps a tool could be written (or already exists) that can deal with such links. Nonetheless leaving it a disambig is sensitive to the practicality that 'laymen' are being introduced to the term knowledge graph by seeing it in google first (as unjust as that is) Fmadd (talk) 23:06, 13 June 2016 (UTC)
- May I note that this as a product name is Title Case, and thus we can have both Knowledge Graph for this page and Knowledge graph for the generic term, with hatnotes on both? --Nat Gertler (talk) 00:58, 14 June 2016 (UTC)
- interesting observation and potential solution, however I think relying on capitalisation for radically different meaning is rather error prone/hazardous. and remember the potential value here for AI, it's not just wether a human can tell.. in the not too distant future we'll be making natural language queries etc based on this dataset (the better the links are, the sooner) Fmadd (talk) 01:20, 14 June 2016 (UTC)
- May I note that this as a product name is Title Case, and thus we can have both Knowledge Graph for this page and Knowledge graph for the generic term, with hatnotes on both? --Nat Gertler (talk) 00:58, 14 June 2016 (UTC)
- IMO I describe practicalities for transition, rather than technicalities; the issue is the accuracy of the 'state' of information in wikipedia at any one time. Perhaps a tool could be written (or already exists) that can deal with such links. Nonetheless leaving it a disambig is sensitive to the practicality that 'laymen' are being introduced to the term knowledge graph by seeing it in google first (as unjust as that is) Fmadd (talk) 23:06, 13 June 2016 (UTC)
Without weighing on whether to use "Google Knowledge Graph" or "Knowledge Graph" (full disclosure, I work on it at Google), I just wanted to point out that I'd like to see at least a few good sources for the claim that the terms "knowledge graph" and "semantic network" are synonymous. Semantic networks are a rather broad term in knowledge representation, whereas the term knowledge graph today seems mostly used for RDF-based graphs. I have to admit that I personally have only a rather vague understanding of what a semantic network really is (and our own article on that is rather bad), but doing research on Semantic Web technologies and RDF for more than a decade, I remember that it was always distinguished from the term semantic network. A simple redirect from knowledge graph to semantic network would sound wrong to me. I would personally prefer a disambiguation over a redirect, which would allow for more flexibility. But, as said, I am obviously a potentially biased voice here, and just wanted to point this out. --denny vrandečić (talk) 22:37, 14 June 2016 (UTC)
- Are you saying that Google Knowledge Graph is RDF? My assumption is that it is not. And after researching just now, it seems that 20 years ago, say, you're right -- there was a sharp distinction between semantic networks and knowledge graphs ([2] "In conceptual graph theory and in semantic network theory the number of relation types is not limited. Whenever some type of relation is needed, this relation will be added. In knowledge graph theory, the types of relation are limited in number, only the eight relations and four frames should be enough to express all semantics. This is the major difference between the two theories.") but in recent years this distinction seems to have gone away ([3] "By knowledge graph we refer to any graph used to represent knowledge about concepts and relations between them. A knowledge graph can be seen as a property graph, a graph whose nodes and edges have properties. Also, knowledge graphs are a superset of multigraphs because they can contain multiple edges between the same pair of nodes.") A good summary of the vagueness and drift of the term is [4]. Perhaps there should be a "Knowledge Graph" section added to the Semantic Network article, since formerly knowledge graphs were a subset of semantic networks and now there is less distinction. Having a separate article for Knowledge Graph (that is, not specific to Google) would require a lot of duplication from the Semantic Network article. Disclosure: I am a developer on an Oracle cloud product that advertises its use of a knowledge graph. Michaelmalak (talk) 11:04, 15 June 2016 (UTC)
- I just now (after some further research) added a paragraph to Semantic network#History about the evolution of the term knowledge graph. If and when there is ever a separate article for Knowledge Graph (in the generic sense), that paragraph can be moved there. But as I stated above, I think because knowledge graphs are a subset of semantic networks (and now nearly synonymous with them) that treating knowledge graphs on the semantic networks article makes the most sense. Now there can be more confidence about either redirecting or disambiguating Knowledge Graph over to the article on Semantic Networks. — Preceding unsigned comment added by Michaelmalak (talk • contribs) 20:17, 15 June 2016 (UTC)
Before we rush to reassign this article, we should reflect on the fact that this article has existed for years, is linked to in a fair number of pages (some through Template:Computable_knowledge, but not all), and that so far the only effort to create an article for the other use of the term seems to be in reaction to the existence of this page. As such, it may well be that this is the topic that is of more current interest, and even if we disallowed separation based purely on capitalization, this page may actually be the reasonable primary at the moment. Additionally, if we are to rename it, we should look into whether the more proper name is "Google Knowledge Graph", or "Knowledge Graph (Google product)". --Nat Gertler (talk) 21:06, 15 June 2016 (UTC)
- I have to agree based on the 2012 hockey stick. [5] I made the WP:BOLD move to spell out the distinction in the Knowledge Graph hat note. I also changed the Knowledge Engine disambiguation page point to Semantic network#History for "Knowledge graph" (lowercase g). — Preceding unsigned comment added by Michaelmalak (talk • contribs) 21:27, 15 June 2016 (UTC)
- "Before we rush to reassign this article, we should reflect on the fact that this article has existed for years," - Thats why I would suggest making this page a disambiguation. Anywhere that links to it gets the error, then can redirect to 'semantic network' or 'google knowledge graph' appropriately.Fmadd (talk) 21:43, 15 June 2016 (UTC)
- Good hockey stick reference, Michaelmalak! That gives some pretty strong suggestion that this Google product is indeed the primary topic at this time. A disambiguation page seems fairly inappropriate when there are only two items and one of them is a clear primary; a hatnote carries things just fine. When there's a primary topic, going to the page will usually get you to the right place and in the minority of cases, the right page is only one click away; with a disambiguation page, going to the page will never get you to the right place, the right place is always one click away. --Nat Gertler (talk) 00:42, 16 June 2016 (UTC) Added: Oh, and in general, we don't rely on the community to fix links we break when we make a move; it's considered appropriate for the person doing the move to correct the links. --Nat Gertler (talk) 00:47, 16 June 2016 (UTC)
- have you seen the context above - there's tonnes of prior research using the term 'knowledge graph' in academia. It's only in *public* use that people associate it with google. Wikipedia surely should be about a neutral POV. Wikipedia is also a work in progress. Future articles may document other research. I have placed a link to this on my own page and 'google knowledge graph' is NOT what I wanted. Fmadd (talk) 20:26, 19 June 2016 (UTC)
- Yes, I've seen the context above... but when the question comes to "when people come to Wikipedia to find out about a 'Knowledge Graph', what are they seeking information on?", I consider online searches to be a far better indicator than what's been written up in academia. (Per WP:PRIMARYTOPIC, "A topic is primary for a term, with respect to usage, if it is highly likely—much more likely than any other topic, and more likely than all the other topics combined—to be the topic sought when a reader searches for that term.") Future developments may change things, yes, but that's always the case when it comes to how disambiguation falls. The minor local politician today may be the President in twenty years. "If there are only two topics to which a given title might refer, and one is the primary topic, then a disambiguation page is not needed—it is sufficient to use a hatnote on the primary topic article, pointing to the other article." --Nat Gertler (talk) 00:40, 20 June 2016 (UTC)
- "a Knowledge Graph is ..... examples of use in commercial products: Google Knowledge Graph, Wolfram Alpha, ..." Just because laymen think google knowledge graph is the only one, it shouldn't stop Wikipedia from educating people about the full use of the term first and foremost. Otherwise companies can get away with claiming general terms (you want anyone looking for "Paint" to get the microsoft product first?)Fmadd (talk) 01:11, 20 June 2016 (UTC)
- I would want users to get MS Paint first if that was most likely to be the topic they were looking for, yes. I have zero problem that Windows takes people to the Microsoft product, rather than to sheet glass portals or the general concept of windows in computer graphic displays. We should be taking them to the topic they actually want, rather than the topic you want them to want. --Nat Gertler (talk) 01:19, 20 June 2016 (UTC)
- wikipedia is here to educate IMO. a compromise might be to mention 'google knowledge graph' early in the text. "a Knowledge Graph is a ..blah blah.. , as found in Wolfram Alpha or Google Knowledge Graph." Fmadd (talk) 03:46, 20 June 2016 (UTC)
- I would want users to get MS Paint first if that was most likely to be the topic they were looking for, yes. I have zero problem that Windows takes people to the Microsoft product, rather than to sheet glass portals or the general concept of windows in computer graphic displays. We should be taking them to the topic they actually want, rather than the topic you want them to want. --Nat Gertler (talk) 01:19, 20 June 2016 (UTC)
- "a Knowledge Graph is ..... examples of use in commercial products: Google Knowledge Graph, Wolfram Alpha, ..." Just because laymen think google knowledge graph is the only one, it shouldn't stop Wikipedia from educating people about the full use of the term first and foremost. Otherwise companies can get away with claiming general terms (you want anyone looking for "Paint" to get the microsoft product first?)Fmadd (talk) 01:11, 20 June 2016 (UTC)
- Yes, I've seen the context above... but when the question comes to "when people come to Wikipedia to find out about a 'Knowledge Graph', what are they seeking information on?", I consider online searches to be a far better indicator than what's been written up in academia. (Per WP:PRIMARYTOPIC, "A topic is primary for a term, with respect to usage, if it is highly likely—much more likely than any other topic, and more likely than all the other topics combined—to be the topic sought when a reader searches for that term.") Future developments may change things, yes, but that's always the case when it comes to how disambiguation falls. The minor local politician today may be the President in twenty years. "If there are only two topics to which a given title might refer, and one is the primary topic, then a disambiguation page is not needed—it is sufficient to use a hatnote on the primary topic article, pointing to the other article." --Nat Gertler (talk) 00:40, 20 June 2016 (UTC)
- have you seen the context above - there's tonnes of prior research using the term 'knowledge graph' in academia. It's only in *public* use that people associate it with google. Wikipedia surely should be about a neutral POV. Wikipedia is also a work in progress. Future articles may document other research. I have placed a link to this on my own page and 'google knowledge graph' is NOT what I wanted. Fmadd (talk) 20:26, 19 June 2016 (UTC)
- Good hockey stick reference, Michaelmalak! That gives some pretty strong suggestion that this Google product is indeed the primary topic at this time. A disambiguation page seems fairly inappropriate when there are only two items and one of them is a clear primary; a hatnote carries things just fine. When there's a primary topic, going to the page will usually get you to the right place and in the minority of cases, the right page is only one click away; with a disambiguation page, going to the page will never get you to the right place, the right place is always one click away. --Nat Gertler (talk) 00:42, 16 June 2016 (UTC) Added: Oh, and in general, we don't rely on the community to fix links we break when we make a move; it's considered appropriate for the person doing the move to correct the links. --Nat Gertler (talk) 00:47, 16 June 2016 (UTC)
I totally agree with Fmadd that Wikipedia is here to educate. This whole discussion about the Knowledge Graph topic should actually be enough motivation to create an article that highlights the different and partially ambiguous usages of the term. For me, it is a pity, that in 2018, Wikipedia still does not explain knowledge graphs in general, but describes the implementation by Google along with a hatnote that refers to other pages. If there is no desire to explain the terminology, I would at least opt for a disambiguation page. However, there would be enough scientific foundation that supports an explanation of the term "Knowledge Graph" itself: Knowledge Graph Refinement: A Survey of Approaches and Evaluation Methods or Towards a Definition of Knowledge Graphs Nordopossum (talk) 14:30, 19 June 2018 (UTC)
Not neutral - this was my reaction, as it looks like Wikipedia is a marketing instrument for Google now. Google has 1. neither coined the word knowledge graph, 2. they came 5-6 years later after Freebase and DBpedia had a knowledge graph in place, 3. nowadays almost everybody has a knowledge graph like Wikimedia has Wikidata and Springer-Nature has SciGraph + 1000 more examples. Google just re-branded and port-portmanteaued Knowledge base into graph and then popularised it. I am suggesting to move this to "Google's Knowledge Graph" which is like DBpedia and Freebase and Wikidata and others and keep the Knowledge graph article neutral by merging knowledge base into it as a history section. Between 2005 and 2008 for example, we were still calling it knowledge base, but were talking about graphs most of the time: https://scholar.google.de/scholar?q=knowledge+graph+dbpedia&hl=en&as_sdt=0%2C5&as_ylo=2005&as_yhi=2008 This only shows the DBpedia papers, but it concerns whole research areas that were talking about knowledge bases back then and now switched to the term "knowledge graph" as popularised by Google. SebastianHellmann (talk) 13:02, 10 July 2019 (UTC)
- And yet, when I refine your scholar search so it's looking for "knowledge graph" DBpedia instead of knowledge graph DBpedia, so that it's looking for knowledge graph as a phrase and not just two words that both occur in the paper, it comes up with only four results... one of which is specifically on Google, and another seeming not to have the phrase in the PDF that comes up. Of the two papers that remain, one is from 2008, and the other, while undated, does not have any references post 2009, and neither has much depth of use of the term. --Nat Gertler (talk) 14:52, 10 July 2019 (UTC)
- This is what I said. It was called "knowledge base" built with "graphs". They are the same thing back then as they are now, we just call them "knowledge graphs", because of the good marketing of Google. Also this Wikipedia article helped a lot. Now there are thousands of knowledge graphs, but why does Google deserve the prime spot here? It is not neutral, all I am saying. There is no such thing as "THE" Knowledge Graph. SebastianHellmann (talk) 19:10, 10 July 2019 (UTC)
- Here is the search for just "knowledge graph" 2000-2010. SebastianHellmann (talk) 19:14, 10 July 2019 (UTC)
- Even the majority of the 2012-and-up articles that mention "Knowledge graph" mention Google, and it's a popular service that has been specifically named that. As such, it seems likely that people who come looking for information on "Knowledge Graph" are looking for information on the Google service, which is how we determine primary topics versus things that need to be disambiguated.--Nat Gertler (talk) 19:35, 10 July 2019 (UTC)
- Below I posted a quote from Ontology (information science) which is written in an unbiased encyclopaedic style and describes knowledge graph much better.
- Even the majority of the 2012-and-up articles that mention "Knowledge graph" mention Google, and it's a popular service that has been specifically named that. As such, it seems likely that people who come looking for information on "Knowledge Graph" are looking for information on the Google service, which is how we determine primary topics versus things that need to be disambiguated.--Nat Gertler (talk) 19:35, 10 July 2019 (UTC)
"Since Google started an initiative called Knowledge Graph, a substantial amount of research has used the phrase knowledge graph as a generalized term. Although there is no clear definition for the term knowledge graph, it is sometimes used as synonym for ontology.[2] One common interpretation is that a knowledge graph represents a collection of interlinked descriptions of entities – real-world objects, events, situations or abstract concepts.[3] Unlike ontologies, knowledge graphs, such as Google's Knowledge Graph, often contain large volumes of factual information with less formal semantics. In some contexts, the term knowledge graph is used to refer to any knowledge base that is represented as a graph."
.
- So I am probably not the only one thinking of the Google Knowledge Graph as an initiative. I also don't know whether it is appropriate to call an ontology a knowledge graph as stated on the ontology page. This seems wrong, because ontology is something else and Google's knowledge graph is also something else. Overall, Google should get credit for the knowledge graph initiative in the second paragraph and the we could have the full shebang like a definition of the term (or definitions) plus a list of knowledge graphs like you have a List_of_datasets_for_machine-learning_research. This would in my opinion more reflect the current state. We can also say that Knowledge Graphs (with "s") was in the Gartner Hype curve without specifically refering to Google, see here https://allegrograph.com/gartner-knowledge-graphs-emerge-in-the-hypecycle/ So it became a non-Google thing, see also here: https://iccl.inf.tu-dresden.de/web/Inproceedings3044/en and Q33002955 SebastianHellmann (talk) 21:32, 10 July 2019 (UTC)
- @NatGertler: Sorry, just saw the move request down below. I think the first reason for declining it is valid. It is based on a lack of better alternatives, i.e. a proper knowledge graph article. So that is the work we have to put in first. I am willing to help. The second reason (search) is declining steadily the more "knowledge graphs" there are. So people looking for "knowledge graph" want to know about knowledge graphs, not the Google Knowledge Graph or more precisely as part of a larger context which has arisen 7 years after the start of Google's initiative. SebastianHellmann (talk) 22:08, 10 July 2019 (UTC)
- :'sokay. And just for keeping track - knowledge graph actually does point elsewhere at this point. It's only Knowledge Graph - capitalized like a Proper Noun, appropriate for a product/service name - that directs here. Now, there are some folks who feel that we shouldn't have articles that differentiate based solely on capitalization; I'm not sure that I agree, but if that were to be fixed now, everything would point here due to a lack of a proper article. And even if we were to have the two articles, if the Google-specific one is the one significantly more in demand, then it's the other one that would get some parenthetical disambiguation, not this one. --Nat Gertler (talk) 22:29, 10 July 2019 (UTC)
- That is another point: Google_Hangouts, Google_Play, Google_Maps, Google_Search, Google_Drive and anything else Google are all correct using the proper title. Google prepends "Google" consistently to all its products. Here not applied SebastianHellmann (talk) 22:40, 10 July 2019 (UTC)
- Having written this, it came to my mind that Knowledge_graph could be a disambiguation page and this page moved to Google_Knowledge_Graph. This is consistent with Google's naming and also appropriate for the current ambiguous situation. SebastianHellmann (talk) 22:46, 10 July 2019 (UTC)
- @NatGertler: Reading the discussion there is ample consensus to act. The immediate move should be:
- Knowledge Graph to Google Knowledge Graph
- Knowledge graph becomes a disambiguation page and then can be extended to a real article accompagnied by a list of knowledge graphs
- Some reasons:
- Google itself is more neutral, if you search for "knowledge graph" a lot of different knowledge graphs pop up, at least for me. Even more dramatically, the Google Knowledge Graph is not even a real thing, it is a component of "Google Search". The German Wikipedia does it right: https://de.wikipedia.org/wiki/Google#Knowledge_Graph as an internal tool of Google. In the first search results of Google there is no hit on the google knowledge graph, except this wikipedia article.
- There is no evidence what users want exactly here. Also the argument to know what users want is a bit difficult to prove and we should ignore it to some extent. I see the main criteria for content on Wikipedia that it is neutral and informative and also general
- There is also no reason to block the move to disambiguation. Personally, I prefer Wikipedia to be up to date and not following a seven year old campaign
- :'sokay. And just for keeping track - knowledge graph actually does point elsewhere at this point. It's only Knowledge Graph - capitalized like a Proper Noun, appropriate for a product/service name - that directs here. Now, there are some folks who feel that we shouldn't have articles that differentiate based solely on capitalization; I'm not sure that I agree, but if that were to be fixed now, everything would point here due to a lack of a proper article. And even if we were to have the two articles, if the Google-specific one is the one significantly more in demand, then it's the other one that would get some parenthetical disambiguation, not this one. --Nat Gertler (talk) 22:29, 10 July 2019 (UTC)
To add a bit background to this discussion: Mike Bergman recently wrote a comprehensive discussion of the history of the term 'knowledge graph'. Note that I might be considered deeply conflicted on this question. --denny vrandečić (talk) 22:03, 10 July 2019 (UTC)
- this reference is fine: https://googleblog.blogspot.com/2012/05/introducing-knowledge-graph-things-not.html as a starting point and then we can take it from there. There is really no technical definition necessary, knowledge graphs can be built in many ways , can be abstracted from the tech and are really usage-driven. I think this is the main point, not the technology (maybe noteworthy to stress it is a "graph"). SebastianHellmann (talk) 22:34, 10 July 2019 (UTC)
- @Denny: Hi Denny, I think I found a good approach to sort the terminology between the different things:
- * Knowledge graphs: are very entity-centric, so they focus primarily on entities and their relations. So the EntityAttributeValue model applies.
- * Knowledge bases are more focused on axioms, like in prologue you formulate axioms or predicate logic statements. This would also generate a knowledge graph, but more implicitly as part of the Universe. They are a bit more similar to databases where you have n-ary relations or tuples and something like foreign keys.
- * Ontologies: These have a strong modelling aspect and regulate Schema, Axiomatic knowledge as well as terminology of the schema like subClassOf/subPropertyOf. I would see them quite distinctive to RDF or property graphs. They are more something you can put on top of a knowledge graph and you can have several to have several views of the same knowledge graph
- * to compare: Wikidata has a total item-centric view and is therefore a knowledge graph, Wikipedia has a lot of wikilinks, which also make up a graph, however the articles are split into entities and conceptual articles, so it is part graph, part ontology (very broadly). DBpedia is a knowledge graph plus ontologies on top.
- * Semantic networks are historically super messy, like wordnet they order terms, not entities. I also found a funny sentence there "In 2012, Google gave their knowledge graph the name Knowledge Graph." but it is very clear that semantic networks are terms, not entities and not axioms.
- I think these could serve as a starting point. Seems to be quite a distinctive criteria. SebastianHellmann (talk) 11:43, 14 July 2019 (UTC)
Which infobox should this article use?
editI want to add an infobox, but which infobox is appropriate?--Qiyue2001 (talk) 09:30, 20 May 2013 (UTC)
Image request
editI would like to request an image example of Knowledge Graph, for instance of Marie Curie, to show what it is in a very simple way. I cannot make an english version myself, as my Google just shows my local non-english version. TGCP (talk) 16:00, 8 April 2014 (UTC)
- Well I can do that.However is it legal? Am I allowed to upload a screenshot of google's proprietary software? LasPo rocks (talk) 00:10, 20 October 2014 (UTC)
- Interesting question :-) and somewhat ironic, given that google Knowledge Graph (KG) is largely scraped off Wikipedia... Let's check, google "da vinci":
- images are the most obvious issue: copyright. But images of da vinci should be public domain by now. They should be credited in the commons-file information page if they are in the screenshot. All images, the da vinci portraits on top, and below the "Artwork" and "People also search for (other portraits)". Can be done.
- text in the KG should be no problem, if it is a Wikipedia article intro text (but again, give credit in the file description)
- data like "Born: April 15, 1452, Vinci, Italy / Died: May 2, 1519, Amboise, France / Full name:..." is free (and taken from google freebase, which scraped from Wikipedia), same for small captions below images ("Mona Lisa 1517")
- search snippets at the left side (outside KG): very short snippets (words!) should be OK, but images there are a problem (copyright). Therefor crop/cut the screenshot off on the left side (offtopic anyway). Also crop the top, no browser.
- Should be OK like that. Don't google Tom Cruise! ;-P --Atlasowa (talk) 12:47, 20 October 2014 (UTC)
- Interesting question :-) and somewhat ironic, given that google Knowledge Graph (KG) is largely scraped off Wikipedia... Let's check, google "da vinci":
Section "Conversational search," — seems POV, OT & bloated.
editPerhaps not strictly Off Topic, but tangential, thus bloated. (The topic is Knowledge Graphs, not search engines (aspects) in general.) Like Amit Singhal's article, that section seems like part of his resume or fan club.
I think it should be eliminated, but I'll not be so bold, and I'll just shorten it.
--71.137.156.36 (talk) 17:41, 24 May 2014 (UTC)Doug Bashford
Curie
editI have removed the Knowledge Graph image that featured a living person being used do illustrate something unrelated to him without his permission (that was why I used my own KG result at first, as I implicitly gave permission) with a Knowledge Graph of a no-longer-living person, one that for some reason had been requested above. --Nat Gertler (talk) 21:29, 7 January 2015 (UTC)
History
editI've removed a newly-added history section, because while it did have a source relating to the term "knowledge graph", there was nothing shown linking it to the Google Knowledge Graph, and the thing being described did not seem to accord with the Google product; it looked like coincidental use of the same term. However, if someone has a reliable source linking them, that would be good to see. --Nat Gertler (talk) 15:42, 5 March 2016 (UTC)
WaPo article about "unsourced" Google knowledge panels
editI recently added a section to this article about "knowledge panels", based largely on a recent report in the Washington Post (WaPo). The WaPo report says that these knowledge panels omit not just what we here at Wikipedia consider to be valid sources, but even omit Wikipedia itself as a source. I don't think the latter critical point would be clear unless we explicitly say so in this Wikipedia article, but this clarification has been reverted. There is no dispute that the cited source repeatedly mentions Wikipedia, and that the cited source uses the word "unsourced" to mean that not even Wikipedia is used as a source. So why not say so in this article? Here's the WaPo cite:
“ | Dewey, Caitlin. "You probably haven’t even noticed Google’s sketchy quest to control the world’s knowledge", Washington Post (May 11, 2016). | ” |
Anythingyouwant (talk) 03:43, 19 May 2016 (UTC)
- That's like saying that if the Post said that someone has no food, we should add "not even pizza" to it. No source means no source. And at the point where they are pointing to the lack of source, they are not emphasizing Wikipedia, so we do not need to add some POV to their statement. --Nat Gertler (talk) 03:46, 19 May 2016 (UTC)
- I'm kind of puzzled by what "POV" you think would be added here, other than the POV that we should write clearly. If WaPo wrote an article saying that a person was not given food, and it's clear from the context that WaPo meant "food" to include liquid as well as solid nourishment, then there would be absolutely nothing wrong with us saying that's what WaPo meant by "food". Same here. Do you dispute that WaPo meant "unsourced" to mean unsourced even to Wikipedia? If a Wikipedia article cited only Wikipedia itself, then it would be considered unsourced by many people, but that's not how WaPo is using the term "unsourced".Anythingyouwant (talk) 04:00, 19 May 2016 (UTC)
- You're treating Wikipedia as a special case when it is not. Wikipedia-citing-Wikipedia-means-its-unsourced is not a special thing about Wikipedia; if Donald J. Trump says "Donald J. Trump says Hillary is lying", that is also unsourced. If the Encyclopedia Britanica or the Washington Post cite themselves, that is also unsourced, but if Wikipedia cites Trump or Trump cites Wikipedia, that is sourced. As we're talking about Google citing something, Wikipedia is not a special case, and and it makes no more sense to say "not even Wikipedia" than to say "not even Boy's Life, or the Windows '98 Manual, or Joe's Hungry Hungry Hippo Fanblog". You're putting in the POV that Wikipedia is a special case, and a diminutive one to boot, as though Wikipedia is the least possible thing one can use as a source - while there are certainly better things that one can use as a source, there are also worse. (It doesn't help matters that the WaPo is describing panels as "sourced" when only portions of the information in that panel is sourced.) --Nat Gertler (talk) 13:45, 23 May 2016 (UTC)
- Okay, thanks for the reply, and you make a good and interesting point. However, please note that the pertinent sentence in this Wikipedia article specifically gives Betty White as an example, and the cited source does specifically discuss Wikipedia in connection with the Betty White knowledge panels: "The one on the left is unsourced; the one on the right is sourced to Wikipedia". So it's not as though I've picked Wikipedia out of thin air, or out of context, nor have I given any misinformation about what the cited source means by "unsourced". So I think your argument boils down to redundancy and superfluosness regarding the phrase "meaning that not even Wikipedia is cited". You say that when Wikipedia cites itself, then of course that's unsourced because whenever someone cites himself it's unsourced. But there's a difference between a Wikipedia article citing itself, and a Wikipedia article citing another Wikipedia article, just like there's a difference between Donald Trump citing himself and Donald Trump citing his sister, and likewise there's a difference between Google citing an unreliable website and Google citing a reliable source. Even if it's arguably redundant, the phrase "meaning that not even Wikipedia is cited" is short and accurate and reliably sourced, so I hope we can keep it, but I won't try to revert or rephrase anymore. It seems like a decent compromise, but it's up to youAnythingyouwant (talk) 17:22, 23 May 2016 (UTC)
- Yes, but the Wapo article does not treat Wikipedia as a special form of sourcing. It doesn't say "it's sourced, albeit to Wikipedia", nor "it's 'sourced,' if you can consider Wikipedia a source." The treatment of Wikipedia as though it's a possible exception isn't called for, and I will remove it, but I will leave the quote you added to the reference and its reference to Wikipedia. --Nat Gertler (talk) 21:41, 23 May 2016 (UTC)
- Thanks for leaving the quote in the footnote. Cheers.Anythingyouwant (talk)}
- Just for the record, I would like to note that the word "unsourced" is sometimes defined as "using information that is not reliable". That's why I supported clarifying what WaPo meant by the word; they did not mean "using information that is not reliable".Anythingyouwant (talk) 18:06, 24 May 2016 (UTC)
- Yes, but the Wapo article does not treat Wikipedia as a special form of sourcing. It doesn't say "it's sourced, albeit to Wikipedia", nor "it's 'sourced,' if you can consider Wikipedia a source." The treatment of Wikipedia as though it's a possible exception isn't called for, and I will remove it, but I will leave the quote you added to the reference and its reference to Wikipedia. --Nat Gertler (talk) 21:41, 23 May 2016 (UTC)
- Okay, thanks for the reply, and you make a good and interesting point. However, please note that the pertinent sentence in this Wikipedia article specifically gives Betty White as an example, and the cited source does specifically discuss Wikipedia in connection with the Betty White knowledge panels: "The one on the left is unsourced; the one on the right is sourced to Wikipedia". So it's not as though I've picked Wikipedia out of thin air, or out of context, nor have I given any misinformation about what the cited source means by "unsourced". So I think your argument boils down to redundancy and superfluosness regarding the phrase "meaning that not even Wikipedia is cited". You say that when Wikipedia cites itself, then of course that's unsourced because whenever someone cites himself it's unsourced. But there's a difference between a Wikipedia article citing itself, and a Wikipedia article citing another Wikipedia article, just like there's a difference between Donald Trump citing himself and Donald Trump citing his sister, and likewise there's a difference between Google citing an unreliable website and Google citing a reliable source. Even if it's arguably redundant, the phrase "meaning that not even Wikipedia is cited" is short and accurate and reliably sourced, so I hope we can keep it, but I won't try to revert or rephrase anymore. It seems like a decent compromise, but it's up to youAnythingyouwant (talk) 17:22, 23 May 2016 (UTC)
- You're treating Wikipedia as a special case when it is not. Wikipedia-citing-Wikipedia-means-its-unsourced is not a special thing about Wikipedia; if Donald J. Trump says "Donald J. Trump says Hillary is lying", that is also unsourced. If the Encyclopedia Britanica or the Washington Post cite themselves, that is also unsourced, but if Wikipedia cites Trump or Trump cites Wikipedia, that is sourced. As we're talking about Google citing something, Wikipedia is not a special case, and and it makes no more sense to say "not even Wikipedia" than to say "not even Boy's Life, or the Windows '98 Manual, or Joe's Hungry Hungry Hippo Fanblog". You're putting in the POV that Wikipedia is a special case, and a diminutive one to boot, as though Wikipedia is the least possible thing one can use as a source - while there are certainly better things that one can use as a source, there are also worse. (It doesn't help matters that the WaPo is describing panels as "sourced" when only portions of the information in that panel is sourced.) --Nat Gertler (talk) 13:45, 23 May 2016 (UTC)
- I'm kind of puzzled by what "POV" you think would be added here, other than the POV that we should write clearly. If WaPo wrote an article saying that a person was not given food, and it's clear from the context that WaPo meant "food" to include liquid as well as solid nourishment, then there would be absolutely nothing wrong with us saying that's what WaPo meant by "food". Same here. Do you dispute that WaPo meant "unsourced" to mean unsourced even to Wikipedia? If a Wikipedia article cited only Wikipedia itself, then it would be considered unsourced by many people, but that's not how WaPo is using the term "unsourced".Anythingyouwant (talk) 04:00, 19 May 2016 (UTC)
KG comparison
editKnowledge Graph Refinement: A Survey of Approaches and Evaluation Methods Heiko Paulheim, Semantic Web 0 (2016) 1–0. ISSN 1570-0844
- Abstract: "In the recent years, different Web knowledge graphs, both free and commercial, have been created. While Google coined the term “Knowledge Graph” in 2012, there are also a few openly available knowledge graphs, with DBpedia, YAGO, and Freebase being among the most prominent ones. Those graphs are often constructed from semi-structured knowledge, such as Wikipedia, or harvested from the web with a combination of statistical and linguistic methods. ..."
Name | Instances | Facts | Types | Relations |
---|---|---|---|---|
DBpedia (English) | 4,806,150 | 176,043,129 | 735 | 2,813 |
YAGO | 4,595,906 | 25,946,870 | 488,469 | 77 |
Freebase | 49,947,845 | 3,041,722,635 | 26,507 | 37,781 |
Wikidata | 15,602,060 | 65,993,797 | 23,157 | 1,673 |
NELL | 2,006,896 | 432,845 | 285 | 425 |
OpenCyc | 118,499 | 2,413,894 | 45,153 | 18,526 |
Google’s Knowledge Graph | 570,000,000 | 18,000,000,000 | 1,500 | 35,000 |
Google’s Knowledge Vault | 45,000,000 | 271,000,000 | 1,100 | 4,469 |
Yahoo! Knowledge Graph | 3,443,743 | 1,391,054,990 | 250 | 800 |
--Atlasowa (talk) 20:22, 19 June 2016 (UTC)
Requested move 20 March 2017
edit- The following is a closed discussion of a requested move. Please do not modify it. Subsequent comments should be made in a new section on the talk page. Editors desiring to contest the closing decision should consider a move review. No further edits should be made to this section.
The result of the move request was: not moved I feel the need to explain since there is only one bolded !vote. The rest of the contributors here reached a rough consensus that 1) the article the draft that was not officially included in the RM, but was the reason for it, should not be moved to the mainspace, and not to the generic title. 2) that Google's implementation of the concept was the primary topic in modern parlance. In line with Wikipedia policy on naming, the page is not moved. (non-admin closure) TonyBallioni (talk) 22:56, 29 March 2017 (UTC)
Knowledge Graph → Knowledge Graph (Google implementation) – This article is only about the Google implementation of knowledge graphs, that's why it should be renamed. Wikipedia is there to eduacate people, so an article with the title "Knowledge Graph" should be about this general term and not about a specific implementation of it. Furthermore there are tonnes of prior research using the therm "knowledge graph" in academia. ERusz (talk) 09:51, 20 March 2017 (UTC)
- See the discussion above under the header "Title". While there are some strong points made, largely it came down to:
- We only put the parenthetical after the title when we need to disambiguate between Wikipedia pages. There is currently no other "knowledge graph" page to disambiguate it from.
- Even were there to be another page, there is reason to believe that the Google product with that name is what most people are looking for when they search for that term, and thus under our guidelines it should be the page without the parenthetical.
- --Nat Gertler (talk) 18:55, 20 March 2017 (UTC)
- I wrote a draft with the title "Knowledge graph" which is about the general term (hasn't been reviewed yet), that's why I made the suggestion to move this page. As I read the discussion above I had the feeling that a disambiguition page is preferred, which is another reason why I suggested moving this page.ERusz (talk) 11:43, 21 March 2017 (UTC)
- I've taken a look at Draft:Knowledge graph, and it convinced me further that this article should stay where it is. For one thing, your article, at least as it stands, wouldn't seem to fit properly into Wikipedia because it's a discussion of a term rather than a thing that the term represents; it indicates that there are varying definitions for the term describing different things. As such, it runs up against the fact that Wikipedia is not a dictionary.
- Even setting that aside, the sources you use discuss the fact that it's the Google use of the term that has made the term much more popular, which suggests that the Google version should be the primary term.
- I will note that knowledge graph, no capitals, has its own page... although at the moment, that page is just a redirect. I would certainly not be adverse to the Title Case version of it being the Google version, and the lower case (or Sentence case, Wikipedia does not distinguish) be a more generic take. --Nat Gertler (talk) 03:46, 22 March 2017 (UTC)
- I wrote a draft with the title "Knowledge graph" which is about the general term (hasn't been reviewed yet), that's why I made the suggestion to move this page. As I read the discussion above I had the feeling that a disambiguition page is preferred, which is another reason why I suggested moving this page.ERusz (talk) 11:43, 21 March 2017 (UTC)
- Oppose. I disagree with the premise of this move. There isn't some sort of "knowledge graphs in general" concept that Google happened to make a single implementation of (and agree with the above that the Draft article isn't really suited for Wikipedia). This is a particular service that Google made and is unique to it. It'd be like moving AngularJS to Angular (Google implementation) because React is a similar framework to Angular. Well, sure, but they're just the names of two separate products. SnowFire (talk) 20:10, 29 March 2017 (UTC)
- Even though I acquiesced to the original move request last year (putting me pretty much on your side), what you stated above is factually incorrect. "knowledge graph" technology dates back to the 1980s as described and cited at Semantic network#History. Google brazenly co-opted this term and made it into their product name Knowledge Graph. It's analogous to Microsoft co-opting the pre-existing term "windows" and naming their product "Windows". And, continuing the analogy, just as "Windows" in Wikipedia takes you to the article on Microsoft Windows, "Knowledge Graph" -- to be consistent -- should take you to the article on Google's Knowledge Graph. This is wrong from the perspective of history and from the perspective of giving academic credit, but Wikipedia, for better or for worse, goes by present-day popular parlance rather than historical record or academic usage. Michaelmalak (talk) 20:51, 29 March 2017 (UTC)
- Were those two words used next to each other before? Yes. The same would be true of many, many other products. Was Google's KG a simple implementation of a pre-existing concept, akin to a company producing brooms for sale merely called "Broom"? There I'd disagree. Sure, CS professors in the 1980s were exploring concepts that would eventually become "knowledge graph", but Google actually built it, and Google's implementation was hardly a simple implementation of instructions from academics in the 80s (barring the ones that Google itself hired, of course). And even then, I'm not sure it'd matter. For an analogous situation, think of various near-future concepts and devices that may or may not come to pass. Suppose someone actually builds one (say, a space elevator or the like), and calls it just the name of the concept. I'm not sure the musings of people 20-30 years earlier (aka us right now) are the "real" topic here; I think the actual implementation is what matters. SnowFire (talk) 22:11, 29 March 2017 (UTC)
- Even though I acquiesced to the original move request last year (putting me pretty much on your side), what you stated above is factually incorrect. "knowledge graph" technology dates back to the 1980s as described and cited at Semantic network#History. Google brazenly co-opted this term and made it into their product name Knowledge Graph. It's analogous to Microsoft co-opting the pre-existing term "windows" and naming their product "Windows". And, continuing the analogy, just as "Windows" in Wikipedia takes you to the article on Microsoft Windows, "Knowledge Graph" -- to be consistent -- should take you to the article on Google's Knowledge Graph. This is wrong from the perspective of history and from the perspective of giving academic credit, but Wikipedia, for better or for worse, goes by present-day popular parlance rather than historical record or academic usage. Michaelmalak (talk) 20:51, 29 March 2017 (UTC)
- The above discussion is preserved as an archive of a requested move. Please do not modify it. Subsequent comments should be made in a new section on this talk page or in a move review. No further edits should be made to this section.
History (2)
editPossibly that part of the current history section would best belong in a more general article on the subject, unrelated to Google, and/or the section could also point to such other article just under its header (via a related/main article link). —PaleoNeonate – 11:33, 10 August 2017 (UTC)
- I moved the section from the history here, mostly because I have trouble understanding it. I am also rather sure that the research being mentioned here did not directly influence Google's Knowledge Graph, but I leave it here for discussion. It would be good if the paragraph could actually also speak about the content, and not only refer to publication dates. For example, what is a Semantic Link Network, what is a self-organised semantic networking approach to render knowledge? What is the ADF? Etc. As it is now it is mostly reference dropping to mostly one specific author:
Before Knowledge Graph and Linked Data, the Semantic Link Network had been systematically studied for creating a self-organised semantic networking approach to render knowledge. The systematic theory and model was published in 2004 [1], and then updated in 2012. [2] This research direction can trace to the definition of the rules for inheritance in 1998 [3] and the Active Document Framework ADF in 2003[4]. Since 2003, it has been developed toward a self-organised semantic networking method [5], [6]. This work is a systematic innovation at the age of the World Wide Web and global social networking. Its purpose and scope are different from that of the semantic net [7]. The rules for reasoning and evolution and automatic discovery of implicit links play an important role in the Semantic Link Network[8][9]. In recent years, it has been developed to support Cyber-Physical-Social Intelligence.[10] It has been used in many applications, including the general summarisation method. [11]
References
- ^ H. Zhuge, The Knowledge Grid, World Scientific Publishing Co. 2004
- ^ H. Zhuge, The Knowledge Grid: Toward Cyber-Physical Society, World Scientific Publishing Co. 2012
- ^ H. Zhuge, Inheritance rules for flexible model retrieval. Decision Support Systems 22(4)(1998)379-390
- ^ H. Zhuge, Active e-document framework ADF: model and tool. Information & Management 41(1): 87-97 (2003)
- ^ H.Zhuge and L.Zheng, Ranking Semantic-linked Network, WWW 2003.
- ^ H. Zhuge, A novel heterogeneous data integration approach for p2p semantic link network. WWW Conference 2004: 334-335
- ^ H.Zhuge, The Semantic Link Network, in The Knowledge Grid: Toward Cyber-Physical Society, World Scientific Publishing Co. 2012.
- ^ H. Zhuge, L. Zheng, N. Zhang and X. Li, An automatic semantic relationships discovery approach. WWW 2004: 278-279.
- ^ H. Zhuge, Communities and Emerging Semantics in Semantic Link Network: Discovery and Learning, IEEE Transactions on Knowledge and Data Engineering, 21(6)(2009)785-799.
- ^ H.Zhuge, Semantic linking through spaces for cyber-physical-socio intelligence: A methodology, Artificial Intelligence, 175(2011)988-1019.
- ^ H. Zhuge, Multi-Dimensional Summarization in Cyber-Physical Society, Morgan Kaufmann, 2016.
- — Preceding unsigned comment added by Denny (talk • contribs) 15:35, 10 August 2017 (UTC)
@PaleoNeonate:, @Shaded0:, someone else - the anonymous editor doesn't seem to be particularly cooperative and does not engage in discussion here. I already removed the paragraph twice, and since I am attached to the topic of the article, I could be regarded as biased. I would appreciate if you or someone else could do an unbiased review of the text and decide whether its inclusion in the current form is warranted. I am having trouble understanding the text itself, and also to see the relevant connection to the topic, but maybe I'm just being obtuse. Cheers, --denny vrandečić (talk) 16:05, 11 August 2017 (UTC)
- @Denny: Done, take a look. I don't think the content adds anything much to this page and I removed it to be a wikilink to the appropriate page if the info indeed has merit of mention. Any more in depth discussion on the theory should belong on the appropriate page. Shaded0 (talk) 16:39, 11 August 2017 (UTC)
- Thanks! --denny vrandečić (talk) 17:17, 11 August 2017 (UTC)
The timeline was in my opinion the most useful part of article, because there's no other place where to access such scattered information about the expansion of the service to new countries, languages and contexts. --Nemo 18:56, 16 August 2018 (UTC)
Explanations from major edit
editHi everybody! As part of my efforts to improve Google Search, I wanted to rewrite this article as well to make sure that I have all the correct facts from articles linked from there. But usually, I update articles by changing or adding one piece of significant information before the next one, with a proper edit summary to explain my actions. But in the cases where I edit old articles, that is sometimes extremely difficult, time-consuming or simply impossible, leading me to make one massive edit, fixing everything at once. But then I don't have enough character space in the edit summary to explain the details, which is why I just thought about the talk page for that purpose. So, explanations, here we go:
- In the "History" section, it opens with information on "Semantic Link Network". I couldn't understand what that actually fully meant, so I copied the title of the source into Google Books, searched for the word "google" and got zero results. The material does not appear to have any direct relevancy to Google, and even if it has some relevancy to the history of the technology, I believe that information can be conveyed even without its mention.
- I moved descriptive information on the appearance of information from the Knowledge Graph, as well as info on the sources and growth of the knowledge base, to a new "Description" section.
- I moved and rewrote the lead's information on Wikipedia traffic drops to a subsection in a new "Criticism" section, which also has another subsection focusing on the criticism of lack of source attribution. I think this gives the page a cleaner look.
- I removed information on Amit Singhal and May 2013 voice searches; the information is not related to Knowledge Graph but rather voice searching, which is covered by Google Search.
- The information on Knowledge Vault apparently has its own article, but the information is very simple. I revised the text here to explain the original 2014 misunderstanding and Vault's functions. I am going to make a redirect of the dedicated Knowledge Vault article so that it links to this article.
- I removed information on DBpedia and Wolfram Alpha from the "History" section, as no sources were given for that text, but DBpedia is linked in the "See also" section.
- I rewrote the lead to summarize the contents of the article.
If you have any questions, feel free to reply to me here :) LocalNet (talk) 15:24, 10 December 2017 (UTC)
Name clarification
editProposal: better knowledge graph (ontology) article --> making this a dab --> moving this page to Google Knowledge Graph.
The Anome initially moved this page to its full title (GKG); which seems even more appropriate now (when there are hundreds of projects that refer to themselves in whole or part as knowledge graphs -- many having little to do with the Google implementation, and more to do with early semantic link networks. denny makes a fair point that this should be a dab page for now rather than a straight redirect; perhaps in 2016 GKG dominated search results for "knowledge graph" but that's no longer true.
Google itself seems to primarily use the whole name now, and we should do the same for that particular private graph. – SJ + 22:58, 29 June 2020 (UTC)
- This is a reasonably popular page in terms of page views, and I suspect that even once the ontology page gets rolling, this will remain WP:PRIMARYTOPIC, best handled with a hat note here pointing to the other use. If, after a while, we're seeing a lot of traffic bounce from here to that other page, that can of course be reevaluated. --Nat Gertler (talk) 00:11, 30 June 2020 (UTC)
Page move to match other products
editAlmost every Google service and project is "Google <capitalized common noun or phrase>" where the shorthand used in internal docs is often just the common noun or phrase. Among Search products, the only one whose full names and articles don't start with a clarifying 'Google' is SafeSearch, which is intercapped and therefore unusual (and trademarkable) -- not a common phrase.
- Google Alerts
- Google Blog Search
- Google Books
- Google Dataset Search
- Google Dictionary
- Google Flights
- Google Finance
- Google Images
- Google Maps
- Google News
- Google Patents
- Google Public Data Explorer
- Google Scholar
- Google Search
- Knowledge Graph
- Google Personalized Search
- SafeSearch
- Google Voice Search
- Google Shopping
- Google Trends
- Google Videos
It seems to me this article should match the others and move to Google Knowledge Graph.
- It matches the naming convention above.
- It avoids confusion about the common term.
- There may be confusion in the tech media, particularly from 2012-16, but not in scholarly works. (cf: related search in Scholar) There were over 1000 articles referencing "knowledge graph" before 2011, and almost all of the >20,000 articles mentioning it now are refer to a generalized knowledge graph, or to the class of knowledge graphs (sometimes including Google Knowledge Graph among the listed instences, but then always written out fully).
There is now another article at knowledge graph about the general concept dating back to the 70s; and a disambiguation page at Knowledge graph (disambiguation). The all-caps Knowledge Graph article can redirect there or to Google Knowledge Graph. – SJ + 22:31, 30 September 2020 (UTC)
Proposed move
editKnowledge Graph → Google Knowledge Graph
- This article is only about Google's implementation of knowledge graphs. It is also Google house style to give products names that are common nouns (like "Inbox"), and they are generally referred to in formal literature by the full name ("Google Inbox"). It is also WP style to use that full name as the article title. We have a lower-cased article on the concept of a knowledge graph, and there is no need for the confusion of two articles with different cases. – SJ + 22:55, 30 September 2020 (UTC)
- @Sj: Please follow WP:RM#CM. Thanks. wjematherplease leave a message... 00:56, 1 October 2020 (UTC)
- You seem to be arguing for using an official name rather than
WP:COMMONWP:COMMONNAME. It looks to me like folks who aren't Google are more likely to use just "Knowledge Graph" than "Google Knowledge Graph". To give a very loose demonstration of this, while Google's Scholar finds 1810 articles that use the term "Google Knowledge Graph", that is a small portion of the 16,400 that used both "google" "knowledge graph", many of which appear to be referring to this product. --Nat Gertler (talk) 02:53, 1 October 2020 (UTC)
@Wjemather: Done, thanks.
@NatGertler: This is the common name, at least in recent years, in research + technical literature and books - including the search results you link to. (It also seems to be the official name, and matches our WP style for other Google products.) Looking at your search, many of the results still refer to the Google Knowledge Graph using its full name. Running a search to exclude those ([7], and looking at the top 10 results: none of them use "Knowledge Graph" to refer to GKG. Two use "Knowledge Graph" to refer to the concept. Three note Google's adoption of a knowledge graph in 2012 popularized the concept, leading to a rise in use elsewhere. The rest don't reference Google's work (despite showing up in this search; it's hard to search effectively for the term 'Google'.).
- Knowledge graph embedding by flexible translation (59 cites) - general concept of a KG, not specific to GKG.
- Challenges and innovations in building a product knowledge graph (15 cites) - refers to the Amazon Product Knowledge Graph and the Google Knowledge Vault.
- Knowledge graph based search system (79 cites) - only reference to Google is to its search.
- Learning a health knowledge graph from electronic medical records (114 cites) - Mainly uses "knowledge graph" as a general term. Research supported by a grant from Google. References the "Google health knowledge graph (GHKG)", later referred to in shorthand as GHKG.
- Enterprise Knowledge Graph: An Introduction (Opening chapter of 2017 book "Exploiting Linked Data and Knowledge Graphs in Large Organisations") - Capitalized "Knowledge Graph" is presented as a "knowledge representation technique" in this book chapter. First section: "1.1 A Brief History of Knowledge Graph". Notes Google's role in popularizing the term.
- One Knowledge Graph to Rule Them All? Analyzing the Differences Between DBpedia, YAGO, Wikidata & co. (46 cites) - various public Knowledge Graphs (KGs) are compared and contrasted. Notes G's role in popularizing the term.
- Knowledge Graph Embeddings with node2vec for Item Recommendation (18 cites) - Refers to a generic knowledge graph used for embedding other knowledge (an increasingly common use of the term since 2016). No reference to Google or GKG.
- Apparatus and method for knowledge graph stabilization (109 cites) - Korean patent, 2013. Refers to stabilizing a generic KG, no reference to Google or GKG.
- Reasoning Web: Logical Foundation of Knowledge Graph Construction and Query Answering (13 cites) - Lecture notes from a summer school. References "the term Knowledge Graph", "the idea of Knowledge Graph", "Test Driven Knowledge Graphs" and "Google Knowledge Graph".
- BioKEEN: a library for learning and evaluating biological knowledge graph embeddings (19 cites) - references Knowledge Graph embeddings [KGEs], no reference to Google or GKG.
I looked through a few more pages of results to find something that came close to using the bare phrase "Knowledge Graph" to refer to GKG, and found this one (which still demonstrates that it is used as a shorthand and not a full name):
- The Graph of Things: A Step Towards The Live Knowledge Graph of Connected Things (1x cites) -- "we create a Live Knowledge Graph to pave the way towards building a ”realtime search engine for the Internet of Things”, which we call the Graph of Things (GoT). Similar to the Knowledge Graph used in search engines like Google and Bing..."
Overall it seems fair to say that the phrase "Knowledge Graph", like the lowercase version, is more often. used to refer to the general concept than to a Google product; and that when the Google product is intended the name is almost always qualified fully, first, as "Google Knowledge Graph" even if it is later referred to by the two-word-phrase as a shorthand. That makes sense: it is a term that is sufficiently generic that to someone even mildly engaged in the field it does not specify which knowledge graph is meant. – SJ + 04:24, 4 October 2020 (UTC)
Requested move 4 October 2020
edit- The following is a closed discussion of a requested move. Please do not modify it. Subsequent comments should be made in a new section on the talk page. Editors desiring to contest the closing decision should consider a move review after discussing it on the closer's talk page. No further edits should be made to this discussion.
The result of the move request was: Moved. Pending technical move request. (non-admin closure) Vpab15 (talk) 19:13, 9 December 2020 (UTC)
Knowledge Graph → Google Knowledge Graph – This article is only about Google's implementation of knowledge graphs. It is also Google house style to give products names that are common nouns (like "Inbox"), and they are generally referred to in formal literature by the full name ("Google Inbox"). It is also WP style to use that full name as the article title. We have a lower-cased article on the concept of a knowledge graph, and there is no need for the confusion of two articles with different cases. – SJ + 03:30, 4 October 2020 (UTC) —Relisting. Interstellarity (talk) 11:04, 3 November 2020 (UTC)
- You seem to be arguing for using an official name rather than
WP:COMMONWP:COMMONNAME. It looks to me like folks who aren't Google are more likely to use just "Knowledge Graph" than "Google Knowledge Graph". To give a very loose demonstration of this, while Google's Scholar finds 1810 articles that use the term "Google Knowledge Graph", that is a small portion of the 16,400 that used both "google" "knowledge graph", many of which appear to be referring to this product. (Copied from earlier discussion.) --Nat Gertler (talk) 04:16, 4 October 2020 (UTC)- I am not, see above. The common use of the phrase, in books and papers, seems most often to refer to the concept, or to a range of implementations, than it does to GKG. Have you found any recent book or scholarly paper that refers to GKG as "Knowledge Graph"? – SJ + 04:27, 4 October 2020 (UTC)
- Absolutely I have. For example, this search will show you 2019 and 2020 usages of "Google's Knowledge Graph", some with that capitalization, indicating the product named Knowledge Graph belonging to the company Google (and that's just in the summary text displayed, as I don't have access to many of the papers themselves.) --Nat Gertler (talk) 04:42, 4 October 2020 (UTC) Adding: And, since books and scholarly papers aren't the only things or even preferred things for judging what is common, here, (directly quoting someone from Google) are, some from the past month. --Nat Gertler (talk) 04:59, 4 October 2020 (UTC)
- I am a big fan of Wikipedia, because it is always up to date. "Knowledge Graphs" was on the Gartner Hype Cycle in 2018. So the "folks that are not google" are likely not looking for the Google Knowledge Graph at all, when looking for Knowledge Graphs. Keeping the article named like this means that Wikipedia remains in the state of 2011, where Google branded the term in its marketing campaign and not 2020, where lot's of people have a knowledge graph and people want to understand what knowledge graphs are. [ SebastianHellmann (talk) 09:23, 5 October 2020 (UTC)
- This page gets about 400 page views per day on weekdays. The lower-cased knowledge graph, which this page even steers folks to if they're not in the right place, gets less than a third as many. So it looks like people who come to this page really are looking for this page. --Nat Gertler (talk) 14:07, 5 October 2020 (UTC)
- Sorry, personally, I find this page views discussion most uninteresting. It's confusing encyclopedic matters with a popularity contest. Before the year 2000, people entering a cafe were asked whether they want milk in their coffee, meaning cow milk. Now in 2020, they get asked whether they want soy or cow milk or whatever. The world just changes. Please note the line: "Dairy milk redirects here. For the line of chocolate products, see Cadbury Dairy Milk." On the milk article. Absolute analogy, no need to even check page views. SebastianHellmann (talk) (talk) 13:42, 6 October 2020 (UTC)
- But page views goes to our guidelines on how we name pages, as it is an indication of what people are looking for - see WP:PRIMARYTOPIC for that sensibility in our guidelines. Our goal is to get people to the information they seek. --Nat Gertler (talk) 13:58, 6 October 2020 (UTC)
- I read WP:PRIMARYTOPIC and also WP:CONCEPTDAB and it says there that these tools are very shaky and no absolute rule. While there is a whole part (more sensible to me) about 'long-term significance'. My argument here is that the world is changing. So you could argue that it is not time yet to change it and rather wait some more months until it gets clearer. Maybe the guidelines need to be clearer as I read them in an opposite manner SebastianHellmann (talk) 14:57, 8 October 2020 (UTC)
- But page views goes to our guidelines on how we name pages, as it is an indication of what people are looking for - see WP:PRIMARYTOPIC for that sensibility in our guidelines. Our goal is to get people to the information they seek. --Nat Gertler (talk) 13:58, 6 October 2020 (UTC)
- Sorry, personally, I find this page views discussion most uninteresting. It's confusing encyclopedic matters with a popularity contest. Before the year 2000, people entering a cafe were asked whether they want milk in their coffee, meaning cow milk. Now in 2020, they get asked whether they want soy or cow milk or whatever. The world just changes. Please note the line: "Dairy milk redirects here. For the line of chocolate products, see Cadbury Dairy Milk." On the milk article. Absolute analogy, no need to even check page views. SebastianHellmann (talk) (talk) 13:42, 6 October 2020 (UTC)
- This page gets about 400 page views per day on weekdays. The lower-cased knowledge graph, which this page even steers folks to if they're not in the right place, gets less than a third as many. So it looks like people who come to this page really are looking for this page. --Nat Gertler (talk) 14:07, 5 October 2020 (UTC)
- I am not, see above. The common use of the phrase, in books and papers, seems most often to refer to the concept, or to a range of implementations, than it does to GKG. Have you found any recent book or scholarly paper that refers to GKG as "Knowledge Graph"? – SJ + 04:27, 4 October 2020 (UTC)
- Do we have enough content for a generic Knowledge graph article? If so, then there should be 2 articles. If there is not enough workable generic content, then this article should be about the Google critter and named accordingly. — GhostInTheMachine talk to me 15:13, 5 October 2020 (UTC)
- There is a generic knowledge graph article at knowledge graph. This is just discussing what should be at Knowledge Graph, with capitals. --Nat Gertler (talk) 16:06, 5 October 2020 (UTC)
- then that makes life very simple. There should not be both Knowledge graph and Knowledge Graph — so this should be renamed to Google Knowledge Graph — GhostInTheMachine talk to me 19:00, 5 October 2020 (UTC)
- Why should there not be both? We have many cases of such differentiation on Wikipedia, having both, say, cat's cradle for the string figure and Cat's Cradle for the novel. --Nat Gertler (talk) 13:48, 6 October 2020 (UTC)
- @GhostInTheMachine: That was my thinking, too. As for Cat's Cradle: that is the full and only title of the Vonnegut novel. The long-form title of Google's product, in their own documents, is Google Knowledge Graph. The long forms of Google product names exist to disambiguate from the general concept and similar tools -- especially those that end in a common noun. Ex: [8], [9], [10] (showing both short- and long-form usage: "Knowledge Graph Search API - Search for matching Google Knowledge Graph entities.") – SJ + 19:01, 8 October 2020 (UTC)
- Why should there not be both? We have many cases of such differentiation on Wikipedia, having both, say, cat's cradle for the string figure and Cat's Cradle for the novel. --Nat Gertler (talk) 13:48, 6 October 2020 (UTC)
- then that makes life very simple. There should not be both Knowledge graph and Knowledge Graph — so this should be renamed to Google Knowledge Graph — GhostInTheMachine talk to me 19:00, 5 October 2020 (UTC)
- There is a generic knowledge graph article at knowledge graph. This is just discussing what should be at Knowledge Graph, with capitals. --Nat Gertler (talk) 16:06, 5 October 2020 (UTC)
- Oppose / comment. Per earlier RM, I'm not sold the current distinction makes sense, or is anything more than a fringe viewpoint. In my opinion, the lowercase knowledge graph article borders on original research by synthesis or a fringe viewpoint. There isn't some general concept of a knowledge graph that Google happened to implement; it is This Specific System used by Google, and some similar systems have become "knowledge graphs" in the same way that other brands of facial tissue are referred to as "Kleenex". If you check Google Scholar, most uses of the term "knowledge graph" don't mean it in the Semantic Web sense, but rather the more general computer science problem of connecting data and attributes in a graph. SnowFire (talk) 06:11, 7 October 2020 (UTC)
- The concept of a knowledge graph indeed existed before Google's implementation. They acquired existing tools and extended them with some of their own. They chose an existing generic noun for the name rather than a new brand such as "Kleenex" -- as they do for most products. – SJ + 19:01, 8 October 2020 (UTC)
- The generic idea, sure, that pre-exists. There's three usages being used and sometimes conflated here. The generic idea of a knowledge graph is so generic that it applies to most of Graph_theory#Computer_science , what could be thousands of usages or products. "Knowledge Graph", the Google creation, was a specific product released in 2012, for meaning #2. And finally, Knowledge_graph#Implementations is "other products that are kind of like Google's Knowledge Graph" for meaning #3, and is the section I consider borderline OR as well as the Kleenex / Facial Tissue example. To my knowledge, people weren't calling DBpedia and Freebase "knowledge graph" before Google rebranded and expanded them as such. Something like Wikidata is clearly meaning #1 of knowledge graph, but that's not necessarily the most relevant descriptor; it's also considered meaning #3 of knowledge graph, "product that is kind of like Google's Knowledge Graph". Anyway, it's a mess. My point is just that Google's "Knowledge Graph" product is not a clean, obvious, implementation of meaning #1 -> meaning #2; it's a realized, separate product that is one specific idea of a huge knowledge graph. It IS very close to meaning #3 - but only trivially so, because Google's KG is essentially defining the term, then. SnowFire (talk) 22:10, 9 October 2020 (UTC)
- The concept of a knowledge graph indeed existed before Google's implementation. They acquired existing tools and extended them with some of their own. They chose an existing generic noun for the name rather than a new brand such as "Kleenex" -- as they do for most products. – SJ + 19:01, 8 October 2020 (UTC)
- Support. The addition of "Google" is common enough, and makes the title far more recognisable. Andrewa (talk) 06:50, 11 October 2020 (UTC)
- Support to increase recognizability and distinguish this topic from the lowercase knowledge graph. (t · c) buidhe 19:36, 12 October 2020 (UTC)
- Oppose Nat Gertler makes lots of good points above. Note also that none of the titles of the sources use the phrase "Google Knowledge Graph". The closest one of them comes is "Google's Knowledge Graph" but we wouldn't normally use a possessive like that to disambiguate. And there doesn't seem to be much need to disambiguate yet. The closest seems to be Bing Knowledge Graph but that redirects to a small section which indicates that that rival is called "Knowledge and Action Graph". Andrew🐉(talk) 19:37, 26 October 2020 (UTC)
- Support per nom, and move Knowledge graph (disambiguation) to Knowledge Graph.ZXCVBNM (TALK) 12:17, 24 November 2020 (UTC)
- Support Some points around GKG not being the only KG. Knowledge graphs have been created -- primarily in academic contexts since the 1980's. Graphs, in a mathematical sense were first envisioned in the 18th century. The term Knowledge Graph (uppercase) is actually trademarked to a company called Diffbot, who have created the largest Knowledge Graph. To points above regarding the levels of page interest in the GKG versus other knowledge graphs, I believe this is primarily attributable to the marketing uses of Google Knowledge Panels as well as featured snippets. These topics are related to the underlying technology of a knowledge graph. But they aren't the primary uses of knowledge graphs, which typically include news monitoring, marketing intelligence, data enrichment for recruiting, providing structured data about relations for machine learning, among other uses. Google's Knowledge Graph is actually not a great example of a widely used KG within enterprise or computing circles as they don't provide any of the underlying data in a way that can be consumed programmatically or otherwise. There are close to 1,000 peer-reviewed articles on Google Scholar mentioning knowledge graphs and not Google. Botley999 (talk) 12:51, 2 December 2020 (UTC)
- The Diffbot material is irrelevant, as we don't name articles based on who owns a trademark, and this article was in place years before they registered their mark. I'm not sure what Scholar search you thought you were linking to, but the one you did just gives all the articles with the words "knowledge" and "graph", not necessarily together, and not the word "google". --Nat Gertler (talk) 13:47, 2 December 2020 (UTC)
- The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.
Free use limitations
editThe recent addition by @Andrew Davidson: which I have since undone runs into a Free Use problem that we've seen before on this page - while we have sufficient free use justification to override the concern about use of the image of Google's page, that page includes a different, presumably-copyrighted item in it: the photograph. This is how we ended up using the Thomas Jefferson example image, as all the images in it are themselves in the public domain. I have removed this new addition. I'm not sure that another example image is needed, but if so, it should be one that does not include separate copyrighted items of concern. --Nat Gertler (talk) 18:34, 26 October 2020 (UTC)
- Tracing back the picture in question, Google's Knowledge Graph seem to have gotten it from Ballotpedia who seem to have gotten it from YouTube who seem to have gotten it from KCCI who don't say where they got it. From pages like this, I get the impression that the image was issued by the candidate's campaign at some point and so it would be valid fair use as a promotional picture. None of those other sites seem to fuss or care about the copyright status because they are sensibly more concerned with getting the job done. Wikipedia is mired in bureaucracy and so, in this case, is being supplanted by another non-profit that Google is choosing to listen to rather than us. Tsk. Andrew🐉(talk) 19:26, 26 October 2020 (UTC)
Discrepancy between text and reference for the current size
edit"By March 2023, this had grown to 800 billion facts on 8 billion entities."
This sentence points as a reference a Google post from May 2020, not March 2023. And it says "500 billion facts about five billion entities".
Most likely the reference is wrong than the numbers. Graph weaver (talk) 11:02, 29 May 2023 (UTC)