The participants were asked 10 questions (tension pairs) regarding their opinion on the state and further development of the Web 3.0 as envisioned by Sir Tim Berners-Lee.

Critics argued that the meaning (semantics) of what we (humans) communicate cannot be easily interpreted (by machines) as we often express our thoughts in "illogical, playfully, misleading, false or nefarious". However, it is not as if we, humans, can always get the point correctly and without an effort, either!

I will try to summarize the report by gathering some of most interesting thoughts about the future of the Semantic Web.

The idea of the Semantic Web is noble, but looks too costly when we take into account such factors as too much variation among people or economic competition.

Some participants question whether Semantic Web could fully emerge in the non-English Web. Those coming from the digital libraries domain note that knowledge organization systems might be more effective than full-blown semantic technologies. Others remind that previous efforts of putting more meta-data on the web pages did not work, and it is hard to talk about RDF, OWL and microformats when well-built HTML is still a rarity.

There are also some concerns about the problem of false data and trust. Some of you might remember that Google did not trust publishers to put correct metadata in the <meta> tags so much that they were not indexing them. False metadata and trust is one of the issues I have been asked about for the last 10 years, ever since I started evangelizing Semantic Web; we have seen metadata abused for so many times that we do not trust that people will play fair with the semantics. That is why the top layers of the semantic layer cake are supposed to prove the correctness of data and reasoning. But these are the top layers, while we are still just half way there, so no wonder many people question whether Semantic Web will reach its full potential by 2020.

Some argued that although developing the Semantic Web is an enormous and costly task, which requires too much coordination and energy to affect the real world, they expect positive results from efforts in creating a semantic web, such as more meaningful, interoperable web content and a better understanding of the Web ecosystem.

Many people believe that Semantic Web will pan out and will take us from the age of information the age of knowledge; this will drive both innovation and exploitation.

Some believe that when Semantic Web is successful no one will call it that way and we will hardly notice its existence. I guess, it is the same as with AI; we tend to say it did not pan out, but at the same time everyday it helps us from behind the scenes. However, in order to make the most of it we need new science (Web Science) to better understand the Semantic Web. In the worst case in 2020 Semantic Web will be "a healthy toddler".

Although there is some resistance for the development of the Semantic Web, there are also strong incentives to cooperate. TBL has a vision that is grander than simply better information retrieval on individual sites; he is going to play important role in the development of the Web over the next decade.

Changes like the Semantic Web are inevitable; however, they should not be driven through arbitrary super-imposed standards, but rather grassroots.

The next generation of the meta-web is inevitable. By 2020 the Web is going to be more semantic than it is today; new solutions will emerge and will continue to conform to their own standards than those proposed by the Semantic Web orthodoxes, like TBL. These standards need to take into account a difficult to harness human element in the process of generating information. Current efforts by Google (Rich Snippets) and Facebook (Open Graph) already show that W3C standards do not always fit and certain changes/adaptations are being proposed. Similar things happened to heavy approaches like SOAP and XML-RPC that were replaced bottom up by REST-ful interfaces. Still the Web has a long way to go before semantic links become the norm.

The Semantic Web might happen but not just yet, not before it finds its killer app, which might not be in the next 10 years

The Semantic Web will require significant behavioral changes that might take longer than a decade. It will take somebody to invent something we all find we need. The next killer app might be a conversational search, especially when we have (will have) the GGG (Giant Global Graph) of interconnected tuples from services like the CIA Fact Book, DBPedia, FOAF, etc.

We will need machines that can preprocess confused and complicated information created by humans in order to create the Semantic Web.

Without conscious Turing-capable computers the Semantic Web looks like a very hard problem to deal with. But when we get there, people might not notice it except to complain about the given search results. Therefore the clear benefits of achieving the Semantic Web need to be obvious to an average user.

The AI component of the Semantic Web, required for machines to understand human, does not offer much hope.

Many people (including myself) are quite cynical about how much can be done by automatically extracting meaning from information; especially when we limit ourselves to the text corpus only. It is like having a kid trying to understand the conversation of his/her parents just by listening to a part of it; sometimes this part is even purposely veiled in phrases the kid *should not* understand. Most likely in 10 years people will still continue to work with unstructured and semi-structured data. The quality of adoption and use of semantic technologies may vary along the way. The Semantic Web might prove useful in certain domains but will rather not evolve to its full potential as envisioned by TBL. Obviously specialized users are more likely to take advantage of the Semantic Web. The Semantic Web will be just like AI - always around the corner in theory and disappointing in practice.

Some think that with information storage becoming more accessible semantics will reside in the application rather than be a part of the web language. I don't think so, we already have "semantics" in applications, and what is really missing is the "links", i.e., the web, component.

If we assume that people who create sites and databases will do that "by the book", i.e., information standards, the major progress towards the Semantic Web will be made; however, plenty of (human) factors are likely to stand in the way. The success measure depends heavily on the definition: some people see the Semantic Web as an optimized design principles for the Web. If you remember our article on the Pedantic Web, it makes a lot of sense to "heal" the Web.

Other respondents noted the role that the Semantic Web standards play in improving data portability and interoperability.

Some argue that by 2020 mechanisms for interoperable computer programs will allow easy and invisible sharing of data across applications. In broad sense the Semantic Web is about the machines understanding humans; but the details of this concept evolve(d) over time along the changes to technology and socioeconomic: when we have more computing power, more data might be better than fragments of high quality meta-data annotated information. Some respondents argue that the Semantic Web is "a way" not the actual "thing" that could be achieved.

The Semantic Web already has had an effect on new applications being delivered; to mention just few: TripIt, TrueKnowledge, Wolfram|Alpha or Twine.

Hence, another success measure spoken about: growing of web applications make effective used of semantic technologies (e.g., more use of knowledge organization systems instead of tags) and ... the number of startups that invent new pieces of the Semantic Web puzzle.

To sum up, let's look at whether the respondents thought that the Semantic Web can be achieved by 2020. 41% of all participants (38% of experts) agreed that by 2020 TBL's vision of the Semantic Web "will have been achieved to a significant degree". While 47% of respondents (52% of experts) argued against, suggesting that by 2020 "users will not have noticed much of a difference".

Kate Ray took the opportunity of having so many experts in the field gathered in one place to record a very impressive video document on Web 3.0. Here is quick overview of of what can be learned from it.

Why ?

We are flooded with massive amounts of information; it gives (would give) us massive potential, but we do not have tools to help us deal with it, yet!

There are almost 10,000 TB of data on the Web; this can be roughly estimated to almost 2000 libraries. David Weinberger notes that old solutions, like Dewey Decimal Classification system (known from the library domain) are no good in classifying content on the Web. Nova Spivack goes even further and claims that Google (or rather the interaction style we use today) is no longer any good in helping us to manage these billions of billions (and counting) Web items. We have to do all the data aggregation and query rewriting in our heads. Sir Tim Berners-Lee (TBL) observes that people understand search as a process that can "parachute" them straight to the answer.

Chris Dixon points out that users overwhelmed with information (options) are less likely to buy anything. Weinberger asks, with all this massive information, should we start tagging our tags (that become unmanageable as well) and filtering our filters?

The conclusion is that in order to integrate, keep update of and filter such a massive (and growing) information space we need to have some structure; and this is where Semantic Web technologies come handy (hopefully).

How ?

TBL says that 20 years ago he showed new ways for how we can deal with information (i.e., by inventing the Web). Now, he wants to encourage us to put our data online. Nova suggests that Web should/will become more like a global database. With a growing number of links between information items expressing relations we get context which is the key step towards inferring the meaning of information.

Critics

Of course not everyone is so optimistic about Web 3.0, even among Web practitioners and experts. For some, the Semantic Web operates in its own unrealistic bubble, and is designed for a idealistic world where everyone is a trained database or information management expert. Since we could not make machines understand us, we now try to "think" (express ourselves) like machines. However, the fundamental question is "Does the World make sense or do we make sense of the World?". Still, the world around us is too ambiguous and there is no single way to express everything; why should there be? It reminds me of the Gödel's incompleteness theorems.

Schism

Just as there is no single way to express things, it looks like there is not single way of thinking about the Semantic Web. David Karger and Abraham Bernstein argue that Semantic Web can be achieved without ontologies: "a little structure goes a long way"; but others say: "Is the Pope a Catholic?"

Future

Everyone agrees that the Web is a huge technological and social success. Nova says that there is always a revolution after evolution; so we now wait for the next killer application to show up. TBL says that the Semantic Web is the platform, just like the Web, so there is no way to tell what should we do with it or where it will get us. He concludes that if the Semantic Web would only realize his vision (or predictions) it would be in jeopardy.

One way to present the vision for Web 3.0 development, actually the Semantic Web to be precise, is to analyze the so-called Semantic Web Layer Cake, which shows a stack of technologies and standards needed to create a fully functional Semantic Web. One of its first version (shown in 2002 by TBL) was as follows:

Semantic Web Layer Cake (2002) Source: http://www.w3.org/2002/Talks/04-sweb/slide12-0.html

The foundations of the Semantic Web are two technologies: URI (Uniform Resource Identifier) and Unicode; they enable us to identify and indicate any content or object (not just online) and describe it in any language. A URI is similar to a URL (Uniform Resource Locator) address, which we deal with every day on the Internet; URIs, however, enable (at least theoretically) to indicate the real-world objects, and thus to describe and to link them to documents on the Internet. Unicode enables one to write with characters from all currently known alphabets.

The second layer is a currently very popular standard called XML, and so-called namespaces associated with it. With XML you can save information so that a machine could process it (but not understand it!). In an upcoming article we will try to explain why introducing the XML layer to the "semantic cake" was not such a good idea.

A third layer defines RDF (Resource Description Framework), a standard that is inherently bound to the Semantic Web. Unlike XML, it allows you to create a (directed) graph in a form processable by machines; based on information stored in these graphs of meaningful relationships (i.e., semantics) a machine can attempt to understand the presented content.

And the next two layers? To assist the process of "understanding" of information stored in RDF, the next two layers define terms used in describing an RDF graph. Yes, you guessed it right: it is all about the infamous "ontologies" (coming up in following articles). RDF Schema can be used to define a simple ontology using a hierarchy of classes and properties; but these classes and properties have nothing to do with object-oriented programming (OOP) apart from sharing the same name. We will discuss this issue in a future article.

The sixth layer, the rules, enables to define (in the formal languages such as RIF) rules for processing of knowledge stored in RDF and ontologies. Together with logic frameworks rules will enable machines to understand the information, rather than merely process it.

The top two layers are necessary to ensure that machines have (some) autonomy in information processing and decision making on behalf of their owners. However, this requires a formal means of delivering evidence (called proof) that the reasoning process was correct; plus a way to determine a level of confidence in the reasoning process (called trust).

The other two elements, digital signatures and encryption, ensure the safety of operations in the Semantic Web.

Phew, a lot of layers, and worse still: all of them the same dry specifications, recommendations, and standards. Fortunately, over the years we managed to change this vision into reality, at least partly.

In recent years, the "Semantic Web Layer Cake" has evolved along with the implementations of the successive layers. Already two years after the initial diagram was published, TBL added an additional layer: queries and XML schemas. There were also the first tool (Annotea), dictionaries (P3P , CC / PP) and standards (Dublin Core, RSS) that implement the bottom layers of the cake:

Another version of the semantic cake (2004) http://www.w3.org/2004/Talks/0412-RDF-functions/slide4-0.html

The current version of the cake contains mainly the specifications that have become W3C standards that implement the different layers: OWL (Web Ontology Language), RIF (Rule Interchange Format), SPARQL (query language for RDF). It is important that ontologies, rules and logic have become a de facto single layer. Further, the role of XML has been limited, because in the Semantic Web we can handle stuff without the restrictions imposed by XML. There is also a new delivery layer: UI (User Interface) and applications. In addition, identification of objects is now possible using the IRI (Internationalized Resource Identifier) standard that extends URI with the ability to use non-standard ASCII characters to encode the object identifier.

Until recently, all the emphasis in the Semantic Web research domain was focused at understanding the meaning of the content by machines. Thanks to the success of Web 2.0, we understood the need and importance of human interaction with computer systems systems utilizing semantic techniques.

The current version of the semantic cake http://www.w3.org/2007/03/layerCake.png

Finally, the cherry on the semantic cake is the 3D version prepared by Benjamin Nowack. What it important here is the new element in the stack of semantic technologies: linked data. They are extremely important for the success of Web 3.0: without semantic, i.e., linked, data (often publicly available) no reasoning mechanisms can do anything.

Semantic cake in 3D, together with related data http://bnode.org/blog/2009/07/08/the-semantic-web-not-a-piece-of-cake

]]> http://www.semanticschool.com/en/2010/04/semantic-web-layer-cake-how-to-digest-the-semantic-web-vision/feed/ 0 http://www.semanticschool.com/en/2010/04/how-old-is-web-3-0/#utm_source=feed&utm_medium=feed&utm_campaign=feed http://www.semanticschool.com/en/2010/04/how-old-is-web-3-0/#comments Fri, 09 Apr 2010 08:37:41 +0000 Sebastian Kruk http://www.semanticschool.com/?p=1109 When I think about adding another number after the term Web I get shivers on my spine. You too? Alright, it was a little difficult - but we somehow got used to the concept of Web 2.0 invented by Tim O'Reilly. But Web 3.0? I hear these voices: People it's time to stop that madness, we've barely started to deploy Web 2.0 solutions in our company. Aren't things moving too fast?

Well, before we get into any "holy" war over the Web 3.0 term, let's try to determine how "old" it is?

1. Web 3.0 does not exist yet, it's just a pipe dream of researchers, and just like artificial intelligence it will never become a mainstream technology. True, there is something in this - we do not stumble upon Web 3.0 sites every day. Twine, digi.me, or even the recent extension of Facebook do not prove anything. Or do they?
2. Web 3.0 is just 6 months old - in March 2009, during the celebrations of the 20th anniversary of the Web, its creator, Sir Tim Berners-Lee, has announced that we already have all the necessary mechanisms and technologies to build Third Generation Internet (or Web 3.0 in the absence of a better term). Well, that's something, but really was there nothing before that?
3. Web 3.0 is really on the Semantic Web, so it should be around 9 years old - in 2001, Sir Tim Berners-Lee, James Handler and Ora Lassila published a famous article in Scientific American. Wait a moment, but if Web 3.0 is as "old" as the Semantic Web, it would be older than the Web 2.0. Something is off here, right?
   I forgot to mention that at some scientific institutes we had tried to achieve what really should be the Semantic Web for far too long; we lost time researching and developing advanced applications (Description Logic or Semantic Web Services). In the meantime the social technologies have become so popular that the Semantic Web had to became more like the Social Semantic Web, and it required a subsequent digit in the label. Well great, now you gonna tell us that Web 3.0 may be even older??
4. Semantic Web already existed in the year 2000. Why? Because that's when the first description of ontologies - in the DAML (The DARPA Agent Markup Language) - was published. True, that's pretty concrete evidence, particularly since DARPA backs it up. Well, but if in the year 2000 we had the first ontology there had to be something earlier!
5. A year earlier, in 1999, Stefan Decker (et al) published the results of research on OntoBroker, which became the foundation for the creation of DAML. In the same year the W3C published the recommendation for the standard RDF, and later the RDF Schema recommendation. And that's it? No!
6. In 1995, DublinCore organization held their first workshop. Anyone who has ever started an interest in semantic technologies struck first on the Dublin Core schema. At that time it was not an ontology as we would call it; the Dublin Core organization, derived from the (digital) library community for many years considered DublinCore an XML rather than an RDF standard. Nevertheless, Dublin Core is still one of the most popular schemata used in and beyond the Semantic Web domain. Well, let's keep going, as it turns out that Web 3.0 is as old as the Web itself! ... How do we know?
7. In 1989 (21 years ago!) Sir Tim Berners-Lee (TBL),  a consultant for CERN at that time, breathed life into something without which we cannot imagine reading this blog - he created the Web. Well, yes, but the Web was in fact a set of  HTML pages and hyperlinks that weave the WWW. And where are the semantics? Well, the Web was supposed to be something more from the start than just a collection of linked pages. Here's the proof: the Web as TBL saw it 21 years ago:

   

   In other words, Sir Tim Berners-Lee was already thinking about the Web as a network of resource connected meaningfully, i.e., with semantics. Surely we can't go back any further in time than that? Actually...

8. In 1969, research on online collaboration and human-computer interaction led by Douglas Engelbart and sponsored by ARPA, lead to DARPANET - the predecessor of the Internet. Wait, but before that we can hardly speak about computers at all ... True, but ...
9. In 1945 Vannevar Bush proposed Memex, a system for cataloging, linking, and managing knowledge. Many people working on semantic technologies, including semantic digital libraries, see the Memex system as a progenitor of current changes in the Web.

Translated by Sebastian Kruk and Jodi Schneider

Therefore, it is a great pleasure for me to let you know about a (free!) book "Linked Data Patterns - A pattern catalog for modeling, publishing, and consuming Linked Data" that describes the issues related to modeling and publishing of linked data. The authors of this book (actually an ebook), Leigh Dodds and Ian Davis, are well-known contributors in the Semantic Web domain.

This ebook has been also published as a PDF and an EPUB. The authors wrote on their blog that this is not yet the final version, and they encourage the readers to send them any (constructive) comments about it.

Since I personally know both authors and their work in the Semantic Web domain, I sincerely encourage you to read this book.

Translated by Sebastian Kruk

Here's a story explaining why you should be interested in the Semantic Web.

Let us imagine that the leaders of Poland, Germany, Russia and the Czech Republic sit at one table to agree on the "historical truth" regarding the Second World War. What problems might they encounter?

1. They speak different languages.
2. Each nation has different "local" and often contradictory truths about that time.
3. The same places and events have different names.
4. The "objective" truth of those times is concealed deeply in the archives.

To address these problems, we want to enable correct and unconstrained exchange of knowledge between different systems.

The first step is to hire interpreters or to agree on a common language for this conversation; it is not hard to imagine that the representatives of these four countries will speak, e.g., English? This way, none of them will feel "cheated" by speaking the language of one of the participants in the meeting.

That is what RDF (Resource Desciption Framework) really is: a knowledge modeling language, which we will present shortly. Why not XML? ...

Second, each party sees the events of 70 years ago in a different way. And describes them using different formats. The same events, places, and words can have quite different connotations.

Therefore, XML is not a good idea - because to validate an XML document we need a common schema; while in RDF each party defines their own ontologies reusing a common vocabulary as often as possible. Submission of information from several sources does not require prior agreement on an ontology (the equivalent of an XML schema for RDF).

Third, besides having different meaning for historical places or events, each party may use different names. What we need is a common vocabulary, which will link various names for the same thing together.

This problem—linking different names for the same thing—can be solved with automation: common identifiers. While there is no way to force a single identifier for the same meaning, it is possible to figure out that different IDs point to the same concept. Rules found in an ontology, such as the inverse functional property, can be used with inference to uniquely identify a concept.

Last, but not least, an important part of the dialogue is to enable access to archives of all parties, without having to enter into separate bilateral agreements, independent visits to individual archives, etc.

Currently, the most important issue related to growing the global Semantic Web is to combine and provide an open access to various sources of knowledge (called Linked Open Data), such as blogs, wikis, open government data, etc. Only through enabling unified access to the global knowledge will Semantic Web really flourish.

This short example addresses a real problem not only on the Internet, but also in the real world. The Semantic Web is a global, distributed system, which (at least partially) helps (or will help) in solving such problems. So no wonder that President Obama's government decided to use Web 3.0 technologies to build the Recovery.gov portal.

In the next article we will describe how the idea of the Semantic Web was born.

These ontologies and other standards arose from digital libraries research. Many of the ideas that formed the foundations for the Semantic Web come directly from the digital libraries domain; many people closely associated with the development of Web 3.0 also have some experience working with digital libraries. So you may be surprised to learn that until only recently these two research domains, i.e., digital libraries and the Semantic Web, could not find a common language.

When we think of a library, two pictures come to mind: a collection of books and a building where these books are stored. However, it is more difficult to define the concept of a digital library:

• Are digital libraries simply directories (available via the Internet) to search  library resources  (called Online Public Access Catalog - OPAC)?
• Or, is a digital library a collection of scanned documents, such as antique books, published on the Internet?
• Or maybe, a digital library is a collection of resources and metadata available through the web browser?
• Finally, perhaps a digital library is a computer system for publishing collections of resources on the Internet?

In addition, digital libraries can be considered both in historical and technological terms. We also take into account the different aspects of their operation: cataloging, information management, information retrieval and user interaction. As you can see the answer to this question (What is a digital library?) is not trivial. One of the European projects on digital libraries, DELOS (now dl.org), delivered a report that attempts answers this and other questions. I will discuss some of them in the following articles in the "Semantic Digital Libraries" series.

As previously mentioned, research on digital libraries delivered a number of standards that are now quite commonly used on the Internet. But until recently the same digital libraries have not really been a part of the Web. They offer services which are often incompatible among themselves not to mention other Internet services. What is more, not so long ago, digital libraries completely rejected social media or any other unsupervised addition of new content or metadata. Additionally, controlled vocabularies (such as thesauri and taxonomies) maintained and used by digital libraries are usually not updated often enough to reflect the current state of knowledge. We should also point out that the user experience provided by digital library systems often differs significantly from that known from other popular web services, like Facebook or Twitter; hence digital libraries often do not meet the needs and expectations of the Internet users, especially of younger generations.

Research on semantic digital libraries aims to achieve better integration between services offered by digital libraries and other services on the Internet, and to improve the user experience provided by digital libraries.

Semantic digital libraries combine results from three research domains:

• Digital libraries deliver knowledge organization systems, dictionaries, thesauri, classification systems, etc.
• The Semantic Web provides solutions for extending current (bibliographic) descriptions and interoperability with other metadata standards and services on the Internet.
• Web 2.0 introduces social descriptions of resources (tagging) and communities of library users.

You may have heard about the initiative called Library 2.0. Some of the ideas regarding semantic digital libraries are similar to the concept of Library 2.0, which is based on 5 elements: open-source, single sign-on (or shared authentication), open standards, integrated OPACs, and of course social media. However, Library 2.0 lacks two essential elements: semantic solutions to ensure extensibility and interoperability, and relations to Web standards and services. Library 2.0 is very popular at the moment, however, in my opinion, we should refrain from saying that digital library systems that allow tagging are semantic digital library systems; there is a whole lot that those systems do not provide.

To sum up this article, I would like to briefly mention a few projects that are strongly related to the idea of semantic digital libraries. In the following weeks I try to describe them a little more closely.

• DuraSpace started last year as a joint initiative of two communities: a popular digital library system DSpace and Fedora Commons providing the infrastructure to build powerful and rich functionality of semantic digital libraries.
• JeromeDL is a project run by DERI Galway and Gdansk University of Technology, with commercial support from Knowledge Hives. It is based on Elvis-DL, a prototype of  a semantic digital library that I created at the Gdansk University of Technology in 2003. Along with JeromeDL we delivered several other projects (within the cluster project Corrib.org), including, among others, the MarcOnt Initiative that provides services, components and ontologies for semantic digital libraries.
• Greenstone is another very popular open source digital library system that enables, among others, effective use of geolocation information.
• SIMILE is a project run by the MIT Libraries; it provides a number of components enabling easy conversion of a classical digital library (or any other web site) into an interactive, feature-rich service based on semantic technologies.
• BRICKS project was conducted under the EU Sixth Framework Program (FP6); it provided a number of components and an infrastructure that enabled social and semantic annotations on library resources and supported exchange of concepts between digital library systems.

You can learn more about Semantic Digital Libraries at the dedicated site of this initiative; you will find there several presentations (including half-day and full-day tutorials) and descriptions of two books on the subject.

Translated by Sebastian Kruk and Jodi Schneider