Re: [Dbpedia-discussion] SKOS property in DBPedia
2008/1/24, Richard Cyganiak: We couldn't find any indication in the SKOS documentation that skos:subject should be used *only* for creative works. I also asked on the SKOS list if this was okay, and the consensus seemed to be that it's a bit strange, but not illegal. Well, there is no domain restriction for skos:subject, so it's 'legal' to relate anything and skos:Concept with skos:subject, but sometimes inappropriate. Let's think the following statement: dbpedia:Tim_Berners-Lee skos:subject dbpedia:Category:Living_people . I don't think it's good idea to say that TimBL's subject (or topic) is Living_people although we can say that TimBL is categorized as Living_people. A person can be a subject of some works, and a person may be interested in some topics, but I can't imagine that a person has a subject or a topic ... Maybe there is a better choice? Do you have a suggestion for another, more appropriate property to use in place of skos:subject? I've tried similar approach that used Wikipedia as PSI of a subject, and used Wikipedia category as basis to categorize these subjects. Since I couldn't find appropriate terms for this purpose, I defined own vocabulary to describe them. http://purl.org/net/ns/wordmap#category can be used to relate DBPedia resource and its category, though the vocabulary is not well known (so far ;-). Or, since DBPedia already defined many terms for own project, it'd be no problem to define one more property for category relationship. best regards, -- @prefix : http://www.kanzaki.com/ns/sig# . :from [:name KANZAKI Masahide; :nick masaka; :email [EMAIL PROTECTED]]. - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
Re: [Dbpedia-discussion] SKOS property in DBPedia
Hmmm, re-reading some of the SKOS docs I get the feeling that skos:subject is indeed appropriate only for documents: http://www.w3.org/TR/swbp-skos-core-guide/#secindexing | These properties [including skos:subject] can be used for subject | indexing of information resources on the web. Here 'subject indexing' | means the same as 'indexing' as defined by Willpower Glossary. The Willpower Glossary says: | indexing: intellectual analysis of the subject matter of a document | to identify the concepts represented in it, and allocation of the | corresponding preferred terms to allow the information to be retrieved So, skos:subject is intended for use on information resources, that is, documents. DBpedia resources in general are not documents. I'm logging this as a bug in the tracker. I think that a new property in the DBpedia namespace is perhaps the simplest solution, e.g. dbpedia:category. Thoughts anyone? Richard On 24 Jan 2008, at 11:06, KANZAKI Masahide wrote: 2008/1/24, Richard Cyganiak: We couldn't find any indication in the SKOS documentation that skos:subject should be used *only* for creative works. I also asked on the SKOS list if this was okay, and the consensus seemed to be that it's a bit strange, but not illegal. Well, there is no domain restriction for skos:subject, so it's 'legal' to relate anything and skos:Concept with skos:subject, but sometimes inappropriate. Let's think the following statement: dbpedia:Tim_Berners-Lee skos:subject dbpedia:Category:Living_people . I don't think it's good idea to say that TimBL's subject (or topic) is Living_people although we can say that TimBL is categorized as Living_people. A person can be a subject of some works, and a person may be interested in some topics, but I can't imagine that a person has a subject or a topic ... Maybe there is a better choice? Do you have a suggestion for another, more appropriate property to use in place of skos:subject? I've tried similar approach that used Wikipedia as PSI of a subject, and used Wikipedia category as basis to categorize these subjects. Since I couldn't find appropriate terms for this purpose, I defined own vocabulary to describe them. http://purl.org/net/ns/wordmap#category can be used to relate DBPedia resource and its category, though the vocabulary is not well known (so far ;-). Or, since DBPedia already defined many terms for own project, it'd be no problem to define one more property for category relationship. best regards, -- @prefix : http://www.kanzaki.com/ns/sig# . :from [:name KANZAKI Masahide; :nick masaka; :email [EMAIL PROTECTED]]. - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
Re: [Dbpedia-discussion] SKOS property in DBPedia
Fred, On 24 Jan 2008, at 13:31, Frederick Giasson wrote: What is a category for DBPedia? Answering to this question will tell you if it is the good thing to do or not. If one is answering that it is a Wikipedia Category, then I will answer that it is not the good thing to do in my opinion. If one ask me why? I would answer that it is become many of the wikipedia categories are classes No. The Wikipedia category system is simply not appropriate as a class hierarchy. That's not a bug; it serves its purpose well, and the Wikipedia community likes it that way. It is essentially a tagging system, where tags themselves can be tagged. See [1] for an in-depth analysis. SKOS is an appropriate vocabulary for modelling the Wikipedia category system. RDFS is not. Hence, we do need a property for indexing arbitrary non-document resources into skos:Concepts. (By the way: Yes, class hierarchies can be created from the Wikipedia category system, e.g. by using cleanup heuristics and combining it with other data sources such as WordNet. Research on this is ongoing in the DBpedia project and elsewhere.) Richard [1] http://arxiv.org/abs/cs/0604036 and in such a case why not defining them as a Class, and not a category (that could be considered a class for some sense of that class)? What if a category is not a class but something else? It is where the cleaning process is starting :) Take care, Fred - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
Re: [Dbpedia-discussion] SKOS property in DBPedia
Hi Masahide, If one ask me why? I would answer that it is become many of the wikipedia categories are classes and in such a case why not defining them as a Class, and not a category (that could be considered a class for some sense of that class)? Wikipedia category is a classification, but not a Class in RDF/OWL sense (in most cases). You just hit the nail :) You said it: in most cases. In fact, Wikipedia categories can be many things: named entities, concepts, relations, (something else?) So, why defining all wikipedia categories as categories (classification purposes) when it is *clearly* not the case? In my point of view, it just make things worse. In case of ambiguity, discard them, don't pollute the dataset with false assertions. In would encourage, everyone on that list, to read and re-read the work of Fabian on Yago: http://www.mpi-inf.mpg.de/~suchanek/publications/yagotr.pdf For example, a category 'Internet_history' (taken from Tim Berners-Lee example) has 'Internet' as its super category. It's maybe OK as classification, but doesn't work as Classes, because 'Internet_history' cannot be a subClassOf 'Internet' (perhaps subClassOf 'History'). I think this would be a relation. There are such categories as 'Tokyo', which could be used as labes of classification, but not even Concepts (or think it like 'Tokyo as concept' ? sounds tricky). Tricky, but it just found the gray area between a subject concept and a named entity I think. I do support to use wikipedia concepts for useful information. Just concern how they should be related to DBPedia resources. What is a Wikipedia concept? I mean, is it different from a concept? How they should be related to DBPEdia resources? With great care. Take care, Fred - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
Re: [Dbpedia-discussion] SKOS property in DBPedia
Hi Richard, No. The Wikipedia category system is simply not appropriate as a class hierarchy. That's not a bug; it serves its purpose well, and the Wikipedia community likes it that way. It is essentially a tagging system, where tags themselves can be tagged. See [1] for an in-depth analysis. Sure, but we are talking about the use of wikipedia categories in dbpedia; I am not questionning the use of categories in WIkipedia. SKOS is an appropriate vocabulary for modelling the Wikipedia category system. RDFS is not. Hence, we do need a property for indexing arbitrary non-document resources into skos:Concepts. RDFS is not in some case, but it is in some other. (By the way: Yes, class hierarchies can be created from the Wikipedia category system, e.g. by using cleanup heuristics and combining it with other data sources such as WordNet. Research on this is ongoing in the DBpedia project and elsewhere.) Exactly what I am talking about. And as I said in my reply to Masahide: should be handled with grat care. Many others? For sure, there are more than 50 papers that talk about that :) In multiple research groups around the globe. The only thing I say is that it is neither black or white, but there are much gray in there. And putting everything in the same bag doesn't help making it more useful. Take care, Fred - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
Re: [Dbpedia-discussion] SKOS property in DBPedia
Hi Richard See also the other thread about deprecation of skos:subject (I suggest to close the current thread and follow-up on that one to avoid parallel discussions) Richard Cyganiak a écrit : Hmmm, re-reading some of the SKOS docs I get the feeling that skos:subject is indeed appropriate only for documents: http://www.w3.org/TR/swbp-skos-core-guide/#secindexing | These properties [including skos:subject] can be used for subject | indexing of information resources on the web. Here 'subject indexing' | means the same as 'indexing' as defined by Willpower Glossary. The Willpower Glossary says: | indexing: intellectual analysis of the subject matter of a document | to identify the concepts represented in it, and allocation of the | corresponding preferred terms to allow the information to be retrieved Indeed, but the genral notion of resource has evolved in the history of the Web from documents to more and more abstract resources. As you know, I belong to people who consider that there is a continuum from physical documents to abstract concepts, and any distinct limit between document, information resource, named entity, concept ... is arbitrary. So, if we want to avoid endless discussions about that, let's assume that Everything is a resource Everything (including concepts themselves) can be indexed by concepts in order to be retrieved A generic mechanism for that should encompass all resources Indexing writers, musicians, buildings by art style, towns and countries by used languages or religions, restaurants by food type etc ... make sense whenever this is intended to retrieve those resources, not to declare classes and attributes. So, using skos:subject for grouping and retrieving DBpedia resources by Wikipedia categories makes perfect sense to me. That said, I agree the example pointed out by Masahide can seem weird. dbpedia:Tim_Berners-Lee skos:subject dbpedia:Category:Living_people ... but, as said in the other thread, not because of the use of skos:subject, but because of the strangeness of the Wikipedia category Living people, which unfortunately tends to be not very reliable, and subject to permanent modification ... That said, it is perfectly functional : it supports queries retrieving all people indexed in this category. And you don't ask more to such a declaration. So, skos:subject is intended for use on information resources, that is, documents. DBpedia resources in general are not documents. See above. Don't put your foot on this slippery slope ... I'm logging this as a bug in the tracker. I think that a new property in the DBpedia namespace is perhaps the simplest solution, e.g. dbpedia:category. Well, I don't think it's a good idea. How will you federate this property with other indexing pointers? Thoughts anyone? You got some :-) Cheers Bernard Richard On 24 Jan 2008, at 11:06, KANZAKI Masahide wrote: 2008/1/24, Richard Cyganiak: We couldn't find any indication in the SKOS documentation that skos:subject should be used *only* for creative works. I also asked on the SKOS list if this was okay, and the consensus seemed to be that it's a bit strange, but not illegal. Well, there is no domain restriction for skos:subject, so it's 'legal' to relate anything and skos:Concept with skos:subject, but sometimes inappropriate. Let's think the following statement: dbpedia:Tim_Berners-Lee skos:subject dbpedia:Category:Living_people . I don't think it's good idea to say that TimBL's subject (or topic) is Living_people although we can say that TimBL is categorized as Living_people. A person can be a subject of some works, and a person may be interested in some topics, but I can't imagine that a person has a subject or a topic ... Maybe there is a better choice? Do you have a suggestion for another, more appropriate property to use in place of skos:subject? I've tried similar approach that used Wikipedia as PSI of a subject, and used Wikipedia category as basis to categorize these subjects. Since I couldn't find appropriate terms for this purpose, I defined own vocabulary to describe them. http://purl.org/net/ns/wordmap#category can be used to relate DBPedia resource and its category, though the vocabulary is not well known (so far ;-). Or, since DBPedia already defined many terms for own project, it'd be no problem to define one more property for category relationship. best regards, -- @prefix : http://www.kanzaki.com/ns/sig# . :from [:name KANZAKI Masahide; :nick masaka; :email [EMAIL PROTECTED]]. - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net
Re: [Dbpedia-discussion] SKOS property in DBPedia
Hi Richard, In fact, Wikipedia categories can be many things: named entities, concepts, relations, (something else?) They certainly all are skos:Concepts. SKOS was created for the purpose of representing exactly that sort of things -- thesauri, taxonomies, and tagging schemes -- in RDF. Yes I know what SKOS is used for. So, why defining all wikipedia categories as categories (classification purposes) when it is *clearly* not the case? Some of the categories might conceivably be modelled as other things as skos:Concepts. That doesn't change the fact that *all* of them can be modelled as skos:Concepts. You said *all*, so you mean that: A Named Entity A Concept A Relation are all skos:Concept? so, each of these things are: An abstract idea or notion; a unit of thought. Could I ask you how you would define a named entity? Also, what about the Wikipedia Category Elvis? I do I agree that there are gray areas between named entities and concepts. However, is it safe to say that *all* named entities are concept? I don't say I am right here, I only ask questions to make things clearer. In my point of view, it just make things worse. In case of ambiguity, discard them, don't pollute the dataset with false assertions. What false assertions? Saying that all named entities are skos:Concept. Everything depends on how you define things In would encourage, everyone on that list, to read and re-read the work of Fabian on Yago: http://www.mpi-inf.mpg.de/~suchanek/publications/yagotr.pdf It is true that a lot of interesting things can be done on top of the Wikipedia categories. Really depends what for. For ontological purposes? I would say no, or at last barely useful (as Fabian would say). For other purposes? Sure it does, we only have to check on Wikipedia :) This doesn't change the fact that there is value in a faithful RDF representation of the Wikipedia category system, warts and all, using an appropriate RDF vocabulary. That's what we have done so far. What is the value you give to it? (I am just asking the question here, I am not saying that there is none; I only want to know what is the value DBPedia have from these categories). What is the added value. Take care, Fred - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion