Raj has just received his annual bonus amount and is planning on buying a new car. He informs his agent for the budget and preference for the Car Make (say Ford, GM and VW). He would like to take a test drive of shortlisted car models coming weekend but wouldn't like to travel more than 5 KMs from his place of residence. In addition, Raj would like to avail of finance options to bridge the gap between his bonus amount and actual car price. The finance options should reflect the minimum interest rates and flexibility of pre-payment with minimum/no penalty charges.
Assuming, all of the above information is available on the internet (which isn't very difficult), with today's web, it would take significant human intervention to coordinate all of the above requirements. In other words, the agent needs to be Raj himself or some other human entity.
Suppose, this agent is a computer program. And the words underlined above serve as the input / decision factors for the program. Then, the program will take Raj's budget amount and prefered car-make as initial inputs, search the websites of the relevant car companies in the region, and come up with an initial shortlist of car models within budget. It can then, check for dealers within 5 KMs radius and available timeslots over the weekend for a test drive. In addition, it can check the car websites for finance options and do the needful.
Similar such programs/agents can be imagined for a variety of scenarios:
- Trying to fix doctor appointment in a given region on a given day for a particular diagnosis
- Looking to buy movie tickets in local theatres within defined distance radius within your budget and choice of film
- Medical research leveraging results from experiments on the given subject in a particular timeline
- R&D opportunities and many more
So, the next question is - is this possible? Yes, and Semantic Web seeks to realize this vision.
The Web as we see today is web of documents which is meant to interpreted by the humans. Semantic web seeks to make the above example interpretation possible by the machines! For this to happen, the web needs to be 'extended' as web of Data. Data which can be represented, understood and infered by software.
Much of the basic standards in this direction have already been established by the W3C but still, there is a lot of ground to be covered.
In terms of the basic building blocks for realizing the semantic web - the most common representation is the one developed by Tim-Berners Lee, provided below:
A detailed description for each of the above areas is provided here. In layman terms, here's what it means:
- IRI (Internationalized Resource Identifier) - This is a generalization of URI (Uniform Resource Identifier). IRI/URI basically serves to uniquely identify any given entity on the web. Only when we have this identification, we can move up the stack for further interpretation. Per our initial example in this post, each of the car make / model will have an IRI associated with it to identify them on the web
- XML (Extended Mark-up Language) - Allows users to 'Tag' their web documents; these are nothing but hidden labels which annotate the web page. For example, a web page detailing specs of VW Polo many have tags like 'VW', 'Polo', 'Hatchback', 'Dealers' etc. However, the XML tags by themselves do not reveal anything about the actual meaning of these tags or structures
- RDF (Resource Description Framework) - The meaning of these tags is expressed by RDF. RDF takes the form of 'Triples' which is very similar to the Subject-Verb-Object of an elementary sentence. For example, VW is a car manufacturer, Polo is a VW car model, XYZ is a dealer of VW Polo etc.
- RDFS (RDF Schema) - This just describes the vocabulary of RDF; it allows for effective creation of hierarchies using classes and subclasses.
- OWL (Web Ontology Language) - Ontology as Web researchers understand is a document or a file that formally defines the relation among terms. This is very important when we are trying to match terms across different databases while aggregating information and thus, arrive at our search results. For example, VW may use the term 'Dealers' to refer to customer facing sale/service points whereas GM may use the terms 'Partners' to refer the same meaning. In order to reconcile the two terms, we need an Ontology which connects 'Dealers' as 'Partners'. OWL (Web Ontology Language) is a contruct which enables this Ontology definition; which typically takes the form of a Taxonomy (Class, Subclasses etc) and Inference rules.
- SPARQL - This is just a RDF query language to retrieve the information for semantic web applications
The remaining layers are not yet standarized but are required to be implemented for realizing the complete benefit of the Semantic web. Much of the remaining layers deal with ascertaining the source of the information (may be using Digital signatures) and the final assimilation into end user interfaces for semantic web applications.
With the realization of the Semantic web, it would become possible to combine information available in different silos and help in coordination/collaboration on a larger scale. Semantic web can lead to generation of new class of tools which can make knowledge sharing between communities much easier and effective. At a big picture-level, Semantic Web can help human knowledge evolve at a faster pace than ever seen before.
Further reading -
0 comments:
Post a Comment