The Semantic Web FAQ

Discussions

News: The Semantic Web FAQ

  1. The Semantic Web FAQ (15 messages)

    Dan Zambonini has created "The Semantic Web: Everything you wanted to know but were too afraid to ask," a FAQ on the Semantic Web. The Semantic Web, "put simply, [is] like installing a huge relational database on the internet, where anyone can add tables to the database, and anyone can add data to a table." The point of the Semantic Web is summarized like this:
    If you've ever used a relational database — such as MySQL or SQL Server — you'll know how useful they are, which is why relational databases are used 'behind the scenes' on most websites today. By storing distinct data-sets that inter-relate (for example, books, authors, and sales), you can very quickly find data that matches certain rules, such as books by a particular author, the top five best selling books, or top five best selling authors. Although each data-set can be maintained separately, they become increasingly powerful as they are joined together. Imagine the possibilities if these data-sets were not just inter-related inside one database, but each database in the world was also inter-related. The authors table from a bookstore could be mapped to a birth records table in a government department (so, for example, you could get top five best selling books by nationality of the author). A historical database of world conflict could then be included to show the top five best selling books by authors who had been born in a country during time of war. And so on. By treating the many large sets of data on the web as a single database, we'll be able to create some incredibly powerful and valuable tools. The current trend of 'mash-ups' goes some way towards exploring this idea (usually mixing the data from only two data sets).
    One of the core technologies behind the Semantic Web might be RDF, which is used as part of RSS 1.0 (the W3C version of RSS). RDF contains an object which has a set of attributes whch have values in a defined format, extensible by namespaces. The Semantic Web is a meme that's been around for a while; it's nice having a FAQ that tries to clear up exactly what it is.
  2. But does the faq make RDF usable?[ Go to top ]

    The faq is nice, but it doesn't solve the issues and problems with the current W3C approach to semantic web and RDF. It's still just as impractical and pointless as it was last year. I love the idea of a semantic web, but the specs coming out of W3C just makes it harder for developers to use. In another 5 years, W3C will still be pitching RDF and semantic web without making an ounce of significant progress. my bias 3 bits. peter
  3. I could not comment on standards themselves but appliance of them seems to be practical enough to build really nice contextual navigation atop of full text search. I recently attended presentation of ORACLEs Secure Enterprise Search combined with Seamark Navigator software and it looks really cool and useful, especially on the intranet. Seamark Navigator uses RDF to define custom reasoning for the navigation. Does somebody have practical experience with Seamark Navigator and SES combo? Is it as good as it looks?
  4. In another 5 years, W3C will still be pitching RDF and semantic web without making an ounce of significant progress.
    check this out: "Cooperation, Competition, and Markets; Or: Why Expressivity Wars are Stupid"
  5. a long way to go[ Go to top ]

    I´ve worked on several semanticweb projects. According to my experience and observations, the semantic web is suffering because: 1) at the current stage there is no benefit 2) there are no usefull (performance + scalability) implementations 3) there is very slow progress The vision of the semantic web as a global and open database is a dream that is not realistic and very unlikely to happen. It seems that this vision is very similar to what LDAP promised. The reason why LDAP has not become a global, distributed and open directory, is that companies simply dont want to open up their data. Another good example is RSS. There is no commercial RSS Feed which contains the actual text. Most feeds only contain links to their webpages. Why is that ? Correct: Ad´s -> Profit and other reasons. There will be no commercial / profit orientated website that will make its content (easily) machine readable. What is the SemanticWeb really about (technical explanation) ? I believe, that the semanticweb will become a distributed rule based programming standard. Currently there is (mainly) XML, where you can describe structure, but you dont have any semantics. This means, if you get a XML file you must first analyze it and translate the "meaning" of the file into a higher level construct before you can express queries. For example suppose you have a social XML file containing persons with friendship relations to each other. If you like to express the query: "show all persons, that are known by my friends." you need to iterate (using transitivity) through all friendship relations. This has to be implemented yourself. But if the document is in OWL Format (a SemanticWeb standart) and the friendship property is defined as transitive, then you dont have to implement anything on your own. You could just query directly and see the results...nice right ? ;) Using the semantic web, you should be able to deliver not only data but semantics (rules) as well, which produce more data (infered data). The true benefit is that the inferencing process can be executed on the client side. Another advantage is that you can keep the original knowledge untouched, while applying rules on it. therefore you are able to switch different rulesets without destroying the original data. Probably the most important point is that it is very simple to express rules. And it is mostly easier then using alternative techniques / technologies to achieve the previous goals. But wait....didnt I say that there is no clear benefit (right at the beginning of my post) ? Currently RDF and OWL offer a limited number of standard rules. The SW becomes usefull only if yo uare able to define "custom rules" (non standard). Which is currently not possible because the according standart (Rule Interchange Format) is still in progress (and does not seem to be completed soon). Another maybe greater problem is that there are no implementations, which could handle custom rules in a high performance and scalable way. I have evaluated most Javabased SemanticWEb frameworks and RuleEngines as well. The results you get are very disappointing. Some people will now disagree, saying that using Frameworks like the Jena API you can process RDF / OWL pretty fast. Yes.....but Jena (like others) have implemented the standartrules using plain Java ;) and not over a RuleEngine (which can process custom rules). Try some custom rules, make micro-benchmarks and judge for yourself. Another issue: In order to express rich rules, you most be able to include functions. Suppose this simple example: if date() is 23-12-2006 => remind_to_buy_presents() date() returns the current date and remind_to_buy_presents() sends a message to your mobile. One solution to this problem is embedding WebServices. But there is no point in making such thoughts, because we dont even have a RuleInterchange format right now. So there is still a lot to do. By the way: marry xmas to all :)
  6. Andreas I agree with your scepticism. But OWL and rule engines are in a different problem space; OWL reasons about knowledge and generates knowledge not actions (side effects). Btw I think that the famous transitive friend relationship is contrived. I don't think there are that many crucial transitive business relationships that merit the choice of a technology like OWL etc.
  7. hi joost
    Andreas But OWL and rule engines are in a different problem space; OWL reasons about knowledge and generates knowledge not actions (side effects)
    OWL Semantics can be implemented in a RuleEngine. With other words: using a Ruleengine (like JBoss Rules or Jess) you can not only trigger funtions but you can also generate new facts (knowledge). cheers, Andreas
  8. OWL Semantics can be implemented in a RuleEngine. With other words: using a Ruleengine (like JBoss Rules or Jess) you can not only trigger funtions but you can also generate new facts (knowledge).
    that's the approach OWLIM and TRREE use to provide OWL reasoning
  9. Thanks Ugly; I had not heard of OWLIM.
  10. Andreas I agree with your scepticism. But OWL and rule engines are in a different problem space; OWL reasons about knowledge and generates knowledge not actions (side effects). Btw I think that the famous transitive friend relationship is contrived. I don't think there are that many crucial transitive business relationships that merit the choice of a technology like OWL etc.
    OWL is an ontology right, so I don't think the spec itself addresses inferencing. If I'm not mistaken ontologies consist of categories + entities of the categories. Like Person is a type and there might be plumber, electrician, carpenter and painter as concrete subclasses of Person. Bob Villa might be an instance of a contractor, or carpenter. Inferencing is quite useful in business cases, like recommendation systems where the engine infers the category a customer belongs to. The customer never says he is a hardcore gamer, but the system might infer that from the individual's purchasing habbits. One might have a rule like. If the person bought more than 10 games the last 12 months, half those games are FPS, and the person owns more than 2 types of console, then they are a hardcore gamer. my bias 2 bits. peter
  11. OWL is an ontology right, so I don't think the spec itself addresses inferencing.
    OWL is not an ontology, it's a formal language for defining ontologies Depending on the concrete OWL subset that you use (Core, Description Logic, Full) you can use different existing reasoners to do the inferencing
  12. OWL is an ontology right, so I don't think the spec itself addresses inferencing.
    OWL is not an ontology, it's a formal language for defining ontologies Depending on the concrete OWL subset that you use (Core, Description Logic, Full) you can use different existing reasoners to do the inferencing
    right, thanks for correcting my typo. I should proof read. the problem I see is the current crop of reasoners perform horribly. the design of many RDF/OWL specific reasoners are bad and poorly thought out. my bias 2 cents peter
  13. Andreas, What I meant was that OWL-DL has certain characteristics (monotonicity) that make it inherently unsuitable to logic programming (rule engines). Peter; reasoning about customers being hardcore gamers is something that rule engines (logic programs) do, not OWL.
  14. Andreas, What I meant was that OWL-DL has certain characteristics (monotonicity) that make it inherently unsuitable to logic programming (rule engines). Peter; reasoning about customers being hardcore gamers is something that rule engines (logic programs) do, not OWL.
    I wouldn't say it's something "rule engines" do. If when you say rule engine you're talking about an expert system or an inference engine, then yes that is what an expert system can do. My point is that inferencing is useful in many business cases. all of this is interesting discussion, but it still doesn't make RDF or OWL practical or useful. OWL by itself does provide the facility to create ontologies, but without a way to determine how "reliable" the ontology is, it's just junk data. the issue of trust has to be addressed for metadata to have practical use. peter
  15. I see a tendency to move away from OWL/RDF as the heart of the semantic services etc. but to see it as just a serialization form. The issue of reliability that you mention is being adressed in the semantic service space by reputation services. That correlates with the lessons learned as embodied in web 2.0.
  16. What does RuleML do if it does not provide a Rule Interchange Format for the web? Cheers Steve T