Ontology in "semantic web"

I have been attempting to keep up to date with up to date topics in computer science, and I keep reading about the "semantic web".

From my understanding, the semantic web means the following:

1)Information on the web is given explicit meaning

2)It would be possible for web services to automatically process and integrate information available on the web.

Simple enough, Im sure there are specifics that are not covered in these points, but I am not focusing on them right now.

I also am aware of the concept of "Ontology" as a method of implementing the semantic web.

This is what Im failing to conceptualize as practical.

Is there a real world, or a practical example of what this would be like? Is there any examples of this curr开发者_JS百科ently being used?

One very practical "semantic" example is the Rich Snippets supported by google. See the website at: http://www.google.com/support/webmasters/bin/answer.py?answer=99170 . By embedding specific markup you can describe things like "hours of operation" for your business to the search engine. I believe that Bing and Yahoo also support this same standard.

In this case the ontology is defined by a schema.

The "Semantic Web" is really a idealisation and conceptualisation; it is a state where all data and information are represented in a way that is understandable and interpretable by machines, so that information can be retrieved precisely according to to one's intent.

The idea of a Semantic Web was proposed (and probably coined) by Tim Berners-Lee in one of his numerous seminal articles: The Semantic Web, Scientific American, May 2001, along with James Hendler and Ora Lassila. Tim founded the Semantic Web Activity in W3C, which endeavours to pursue this idealisation. Its homepage provides a brief introduction as well as several important references, shall you want to research into this topic in more depth.

Miller and Swick also wrote an introductory article on "An Overview of W3C Semantic Web Activity" in 2003. I found this article a good entry point to understanding what "Semantic Web" means.

It's a web of data. All machine readable. Here's an example, but not a real implementation. http://www.youtube.com/watch?v=RNJl9EEcsoE

Semantic web is a web that is connected in such a way that one point leads to another relevant point.In semantic web every image or data will have a meaning and will not be based on key words.

http://www.business-science-articles.com/science/articles/computer/601-semantic-web

Very brief overview can be found here. Real world examples are out there, and there are plenty of them.

While Semantic Web was originally meant for the web, it is more useful than that. For example, it can be used to discover and build knowledge bases (expressed as ontologies). In my work context we are looking into using it for dealing with changing aspects of domain analysis, and since our perception about the domain changes (or is not yet clear at the moment), it is easier to express knowledge and some dependencies with ontologies.

The alternative could be using standard software engineering techniques, relational databases, diagrams, etc., but again, in our context (scientific and shared knowledge) it makes sense to use semantic web concepts.

The easiest way to think Semantic web is to see how it can be applied to real world scenarios.

Let's take example sentences: 1. John Smith is a professor at MIT. 2. Prof. Smith is also dean of CS.

We, humans, can easily try to relate the above two sentences. For machines, it's not possible. why? because they don't know whether John is a car,person or a street name. And whether Smith is a last name or the job title.

But what if we have following additional information somewhere in the web:

1. 'John Smith' is a person. a. John is first name b. Smith is last name c. unique address on web is http://www.example.com//person/uid/9087809812
2. 'MIT' is university
3. 'professor' is position at university
4. 'Prof.' is same as 'Professor'
5. 'Dean' is position
6. 'CS' is acronym for 'computer science'

Now, if we have additional information (often called metadata) for each entity like above and if they can be dereferenced (i.e. if we can look up using corresponding unique address), machines can relate entities on their own. This is the vision of semantic web- to have meta information about every possible entities and create meta information following certain controlled vocabularies and logics referred to as 'ontology'.

One way to look at ontologies (very simplistic view mind you) is to just focus on the subject-predicate-triples found in ontologies themselves and then refer to some of the already published ontologies out there (many of them from the natural sciences.)

YMMV, but at least for me, the idea of the semantic web and ontologies didn't click on me until I went back to my school years and found an analogy in Prolog's rules and facts.

Using trivial examples, say you want to express information as relations in subject-predicate-object triples (disease "X" caused-by agent "Y", legal precedent "A" related-to some law "B" and so on.) This is not possible by just using keywords. So you build relations to express information, and then those relations are implemented in some machine-readable format.

Then, at least so the idea goes, these relations can be enumerable. They can be processed and people can ask questions such as: is X related to Y by relation Z? What is related to X? What is related via relation Z? You can infer new knowledge this way, specially if one uses a well-established ontology (as in the link I previously provided)...

... or so does the theory goes. It is a good concept, one that people are already putting to good use, and I think we are moving in the right direction. Time will tell if the idea is universally applicable in a practical manner.