According to Nova, the Semantic Web is about connecting „everything“, i.e., concepts, things, words, etc. more than a social graph. Saying that “Joe” is of type “Person” and “Palo Alto” is of type “city”. This is not just a hyperlink.
Web 3.0 starts a new decade, the 3rd decade of the Web. It is about enriching the structure of the web, transforming the web form something that’s more like a fileserver to something that’s more like a database.
The 1st era was that of the PC: focused on the frontend. The Web 1.0 focused on the backend; Web 2.0 focuses on the frontend again (AJAX, etc.); Web 3.0 will be backend again. Like a pendulum that swings back. Web 3.0 is a fundamental upgrade to the infrastructure of the web.
Web 4.0 (the fourth decade) will be smarter interfaces and smarter tools, smarter user experience.
According to Nova, there are five approaches for bringing intelligence to the web:
Tagging: very easy to do (which is a pro and a con). Tags at the end of the day are meaningless, e.g. the same person will tag things differently in different situation. One needs a large data set for making sense out of tags.
Statistical approach: what e.g. google does, very stable, does not add semantics
Linguistic approach: trying to actually understanding the meaning of text; is very computationally intensive, hard to scale, etc.
Semantic Web approach: set of open standards by the W3C, i.e., using meta-data to describe the meaning of data; the use of meta data is open to all applications; needs tools, needs someone who creates the metadata, …
Artificial intelligence: software that really thinks, it will take at least some more decades, scalability is a big issue, etc. AI will be huge in Web 4.0. Cycorp is working on such an approach for the last 15 years or so.
The question is whether you make software smarter or data smarter (or both). The Semantic Web is a kind of compromise.
There are two types of approaches to achieve this:
Everyone is going to manually create semantic web content (bottom up).
Today, we generate RDF, OWL, etc. (semi-)automatically using tools (top down).
Sun says: The network is the computer. Nova says: The Web is the database!
Smart data is data that carries what is needed to make sense of it. The data is self-describing. This allows you to write “dumb software”, i.e., with a general piece of software that you put at medical data will allow you to give medical advice.
We are not trying to replace human intelligence. Machines are good at number crunching; this is what they should do. Humans are intelligent, we need to assist the humans with machines, i.e., computers in the semantic web.
Just-in-time-data is another concept. The semantic web provides the basis for it by using ontologies (like schemas in DBMSs) to achieve that.
The WebOS is coming. The WebOS needs a file system, the semantic web is a candidate for that. With the semantic web we add an open database layer on top of today’s web.
Standards are important: RDF, OWL (built on RDF), SPARQL (SQL for RDF), SWRL (a rule language), GRDDL -> see the W3C site for details.
“DataWeb” as better term for Semantic Web; with triple stores as new types of databases (based on e.g. relational DBMSs) to manage (the large) lists of triples. Scales better on the web.
Geo-names, music-names, sioc, etc. and many more ontologies and connected data do already exist.
Right now still in early adoption of this period. 007-2009 as first wave, a couple of million end-users in their day-to-day- live. twine, freebase, etc.
FOAF; SIOC; are good places to start.
Does the semantic web offer any new business models? The semantic web does not introduce any new business models (in addition to the existing ones such as search, advertise, media & content, …) but it makes things smarter. But the semantic web allows new players to come into the market.
Openness? if your business is about being evil, will you do better with the semantic web? It will be more difficult for you, because the semantic web will generate data that is self-descriptive; i.e. lock-in strategies will be harder to achieve.
I have heard in the year 2000 about the semantic web. Now it’s 2008 and it is still not there. Will it – and when – take off? We often forget that there was 15 years of research in hypertext, protocols, etc. before we could use the web in the early 90ies. We often forget about that. The same (and probably a longer period) will be necessary for the semantic web.
How do you address global definitions? Good question! The semantic web does not require a global definition. The semantic web was designed for disagreement. Therefore, the problem is that there may be many definitions. Hence, the standards were built to include a mechanism to create mappings to define equivalences, etc. The winners will be the ones that have the content that lead the definitions of the standards AND have the data.
Another issue related to this is that all my preferences, etc. data is typically stored with the site. Ideally, I would like to take my comments, etc. with me and be independent of the provider. Well, in an ideal world … (“In an ideal world, the policemen would be English, the car mechanics German, the cooks would be French, the bankers Swiss, and the lovers Italian. But in the real world, we must not forget that there are also policemen that are German, car mechanics that are French, cooks that are English, bankers that are Italian and lovers that are Swiss.”)