Kingsley Idehen's Blog Data Space

As is often the case these days, it's much easier to drop a blog post than it is to make a simple comment in an "old media" style data space :-(

My use of "old media" implies: a place that still seeks subscriber data (no OpenID etc..), for the umpteenth time, as the toll fee for discourse development and participation on the Web.

Anyway, here is what I attempted to post as a comment to Dan Grigorovici's post titled: Where is the Semantic Web Killer App?

Dan,

An intriguing post to say the least :-)

"Linked Data" and "Semantic Web" aren't synonymous, they are simply connected, infrastructure DNA-wise. You can have "Semantic Web" style graphs (i.e RDF Data) and not have "Linked Data" as per Linked Data deployment tenets and best practices, a very important point.

I've stated repeatedly, the "Linked Data" emphasis has more to do with focusing on a point of crystallization within the larger "Semantic Web" vision, so here is a quick recap:

What is Linked Data?

A term coined by TimBL that describes an application of HTTP to the time-tested process of "Data Access by Reference". "Linked Data" adds vital items to the "Data Access by Reference" pattern that have been erstwhile unattainable:

• The use of a Data Source Naming scoped to Database / Data Container Records as opposed to Tables, Views, Stored Procedures, Databases, and other Record Container tuple collections. Example: in ODBC / JDBC, a Data Source Name's scope stops at the Table / View level. In the Linked Data realm you get an added layer of granularity due to record level name scope
• Incorporation of HTTP into the Data Source Naming scheme, which injects the expanse of the Web into the Data Access Range of the Data Source Name (i.e. a Named Record); so you can reference a record's description directly via HTTP which is simply a major deal (to put things mildly).

So we have HTTP based URIs as the Data Sources Names for a "Linked Data Web" i.e a Web of inter-connected Data Source Names that de-emphasize the importance of their host containers (Compound Documents / Information Resources).

The business case or value proposition of "Linked Data" is synonymous with the value proposition of data access technologies such as ODBC, JDBC. ADO.NET, OLE-DB, XMLA, and others (enterprise or consumer) in relation to the Individual and Enterprise pursuit of agility; in a realm where data is growing exponentially, and the maximum processing time in a single day remains 24 hrs. Data Access & Data Integration are timeless challenges due to the following constants:

• Structured Data Schema Heterogeneity - we will always model the same things differently
• Dirtiness of Data within Structured Data Containers - we are error prone due to laziness / sloppiness, time constraints, and the inherent limitation of our DNA based CPUs when dealing with large volumes of data.

Note: The line between the Enterprise & Individuals continue to blur by the second, this is something I covered during my Linked Data Planet keynote, which is like most things I put on the Web (via this blog data space), is a live and practical demonstration of the virtues of Linked Data courtesy of RDFa, the Bibliographic Ontology, and dereferencable URIs (i.e. HTTP based Data Source Names for Documents and the Entities they host).

Related

• Linked Data FAQ - by Mike Bergman
• The Planetary Computer from Wired Magazine - which is basically the effect of Linked Data under a different label (note to Wired: "Tired of old Media repetitive Registrations" when seeking to make comments in the OpenID era!).

Metcalfe’s law states that the value of a telecommunications network is proportional to the square of the number of users of the system (n²), where the linkages between users (nodes) exist by definition. For information bases, the data objects are the nodes. Linked Data works to add the connections between the nodes.

I would tweak of the law modification expressed in Mike Bergman's post which states:

the value of a Linked Data network is proportional to the square of the number of links between the data objects.

Channeling Einstein

How about Einstein's E=mc²? I am talking Energy (vitality) and Network Mesh equivalence, where "E" is for Energy, "m" for Mesh (i.e. Network Mesh where each node contains sub-particles that are themselves network meshes all endowed with typed links and weightings), and "c" is for computer processing speed (processing speed is growing exponentially!). When you beam queries down a context rich mesh (a giant global graph comprised of named and dereferencable data sources), especially a mesh to which we are all connected, what do you get? Infrastructure for generating an unbelievable amount of intellectual energy (the result of exploding the sub-data-graphs within graph nodes) that is much better equipped to handle current and future challenges. Even better, we end up making constructive use of Einstein's findings (remember, we built a bomb the first time around!). TimBL articulates this fundamental value of the Web in slightly different language, but at the core, this is the essence of the Web as I believe he envisioned; the ability to connect us all in such a way that we exploit our collective manpower and knowledge constructively and unobtrusively, en route to making the world a much better place :-)

Note: None of this in incongruent with being compensated (i.e. making money) for contributing tangible value into, or around, the Mesh we know as the Web :-)

Related

• Business at the Speed of Thought - by Bill Gates
• Blink - by Malcolm Gladwell

Mike Bergman has just published a nice Linked Data FAQ aimed at Enterprise audiences. His post draws on a collection of questions collated from a plethora of interactions with Enterprise oriented folks during last week's Linked Data Planet conference.

Enjoy!

Stumbled across a nice post titled: What do people have against URLs?. My answer: Everything, if they don't understand the inherent power of URLs when incorporated into the "Data Source Naming" mechanism of the Web called: URIs :-)

URIs are simple to use i.e you simply click on them via a user agents UI. However, URLs when incorporated into Data Source Naming en route to constructing HTTP based Identifiers, that deliver HTTP based pointers to the location / address of a Resource Descriptions, another matter.

I touched on this issue in my Linked Data Planet keynote last week, and I must say, it did set off a light.

I believe, we can only get the broader Web community to comprehend the utility of URIs (Web Data Source Names) by exposing said utility via the Web's Universal Client (Web Browser). For instance, how do URN based Identity / Naming schemes help in a world dominated by Web Browsers that only grok "http://"? From my vantage point, the practical solution is for data providers who already have "doi", "lsid" and other Handle based Identifiers in place, to embark upon http-to-native-naming-scheme-proxying.

In my usual "dog-fooding" and "practice what you preach" fashion, this is exactly what we do in the new Linked Data Web extension that we've decided to reveal to the public (albeit late beta). Thus, when you use an existing browser to view pages with "lsid" or "doi" URNs, you still enjoy the utility of getting at the "Raw Linked Data Sources" that these names expose.

I've finally found a second to drop a note about my keynote.

The keynote: Creating, Deploying, and Exploiting Linked Data, sought to achieve the fundamental goal of: Demystify the concept of "Linked Data" using anecdotal material that resonates with enterprise decision makers.

To my pleasure, 90% of the audience members confirmed familiarization with the "Data Source Name" concept of Open Database Connectivity (ODBC). Thus, all I had to do was map "Linked Data" to ODBC, and then unveil the fundamental add-ons that "Linked Data" delivers:

• The ability to give database records names (Identifiers)
• The use of HTTP in the database record naming mechanism - which expands a named database record's reference scope via the expanse of the Web (i.e HTTP based Identifiers called URIs).

I believe a majority of attendees came to realize that the combination above injects a new Web interaction dynamic: access to "Subject matter Concepts" and Named Entities contained within a page via HTTP base Data Source Names (URIs).

BTW - My presentation is a Linked Data Space in it's own right courtesy of the Bibliographic Ontology (which provides slide show modeling) and RDFa that allows me to embed annotations into my Slidy based presentation :-)

Related

• PowerPoint version of Presentation
• Slideshare hosted version
• Authorstream hosted version
• Google Docs hosted version

Yikes! I've just discovered that the final part of the semanticweb.com's interview with Jim Hendler and I, includes critical paragraphs that omit my example links :-( As you can imagine, this is a quite excruciating, bearing in mind that "Literals" are of marginal value in a Linked Data world.

Anyway, thanks to the Blogosphere, I can attempt to fix this problem myself -- via this post :-)

Q. If you wanted to provide a bewildered but still curious novice a public example of Linked Data at work in their everyday life, what would it be?

Kingsley Idehen: Any one of the following:

My Linking Open Data community Profile Page - the Linked Data integration is exposed via the "Explore Data" Tab My Linked Data Space - viewed via OpenLink's AJAR (Asynchronous Javascript and RDF) based Linked Data Brower My Events Calendar Tag Cloud - a Linked Data view of my Calendar Space using an RDF-aware browser In all cases, you have the ability to explore my data spaces by simply clicking on the links, which on the surface appear to be standard hypertext links, although in reality you are dealing with hyperdata links (i.e., links to entities that result in the generation of entity description pages that expose entity properties via hyperdata links). Thus, you have a single page that describes me in a very rich way since it encompasses all data associated with me, covering: personal profile, blog posts, bookmarks, tag clouds, social networks etc.

Q. What would you show the CEO or CTO of a company outside the tech industry?

Kingsley Idehen: A link to the Entity ALFKI, from the popular Northwind Database associated with Microsoft Access and SQL Server database installations. This particular link exposes a typical enterprise data space (orders, customers, employees, suppliers ...) in a single page. The hyperdata links represent intricate data relationships common to most business systems that will ultimately seek to repurpose existing legacy data sources and SOA services as Linked Data. Alternatively, I would show the same links via the Zitgist Data Viewer (another Linked Data-aware browser). In both cases, I am exploiting direct access to entities via HTTP due to the protocols incorporation into the Data Source Naming scheme.

The build up to Linked Data Planet continues... Here is semanticweb.com's interview with Jim Hendler and *I* titled: Linked Data Leaders - The Semantic Web is Here.

As I start my countdown to the upcoming Linked Data Planet conference, here is the first of a series of posts geared towards showcasing practical use of the burgeoning Linked Data Web.

First up, the Library of Congress, take a look at the following pages which are "Human" and machine based "User Agent" friendly:

• Science
• Humanities
• Mathematics
• Cataloging
• World Wide Web

Key point: The pages above are served up in line with Linked Data deployment and publishing tenets espoused by the Linking Open Data Community (LOD) which include (in my preferred terminology):

• Giving "Names" to things you observe (aka Data Source Names or "DSNs" for short)
• Use HTTP URLs in your data source naming scheme so that "access by reference" to your data sources exploits the expanse of the HTTP driven Web i.e make your DSNs "Linked Data Source Names" (LDNS)
• Remember that Documents / Pages are compound in nature, and they aren't the only data sources we would want to name; a document's LDSN must be distinct from the LDSNs used for the subject matter concepts and/or named entities associated with a document
• Use the RDF Data Model to express structure within your data source(s)
• Use LDSNs when constructing statements/claims/assertions/records (triples) inside your structured data sources
• When publishing Web Pages related to your data sources; use at least one of the following to methods to guide user agents to data sources associated with your published page; the HTML LINK tag, RDFa, GRDDL, or Content Negotiation.

The items above are features that users and decision makers should start to hone into when seeking, and evaluating, platforms that facilitate cost-effective exploitation of the Linked Data Web.

I just stumbled across a post titled: Why Reasoning Matters: Consistency Checking from Clark and Parsia

As you can see from my recent post about how we've started the process of inoculating DBpedia against the potential dangers of "contextual incoherence", we are entering a newer era in the Semantic Web's evolution. My post and the one from Clark & Parsia both touch different aspects of the "Data Dictionary" for the Semantic Web issue.

Note: in my universe of discourse, a Data Dictionary manifests when the constraints and class hierarchies defined in an ontology (e.g. a web accessible shared ontology) are functionally bound to a data manager. Interestingly the binding can take the following forms:

• Engine Hosted - which is what you get with Virtuoso's in-built Inference Engine
• External - which is what you get when the Inference Engine is a distinct component from the data manager (example: Owlgres which can sit in front of 3rd party SPARQL endpoints via ARQ)

The classification terminology I use above is very much off-the-cuff, its sole purpose is architectural distinction.

Anyway, it's really nice to see that we are entering an era re. the Semantic Web vision, where the virtues of reasoning are getting simpler to demonstrate and articulate.

In a nutshell, the point-point data integration era is coming to an end! The era of intelligent ontology based enterprise data integration is nigh!

Of course, there is much more to come on the practical utility front, so stay tuned as we work our way through the DBpedia inoculation program.

The current live instance of DBpedia has just received dose #1 of a series of planned "Context" oriented booster shots. These shots seek to to protect DBpedia from contextual incoherence as it grows in data set expanse and popularity. Dose #1 (vaccine label: Yago) equips DBpedia with a functional (albeit non exclusive) Data Dictionary component courtesy of the Yago Class Hierarchy .

When the DBpedia & Yago integration took place last year (around WWW2007, Banff) there was a little, but costly omission that occurred: nobody sought to load the Yago Class Hierarchy into the Virtuoso's Inference Engine :-(

Anyway, the Class Hierarchy has now been loaded into the Virtuoso's inference engine (as Virtuoso Inference Rules) and the following queries are now feasible using the live Virtuoso based DBpedia instance hosted by OpenLink Software:

define input:inference 'http://dbpedia.org/resource/inference/rules/yago#'
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX dbpedia: <http://dbpedia.org/property/>
PREFIX yago: <http://dbpedia.org/class/yago/>

SELECT ?s
FROM <http://dbpedia.org>
WHERE {
?s a yago:Publication106589574 .
?s dbpedia:name "The Lord of the Rings"@en .
}

-- Variant of query with Virtuoso's Full Text Index extension: bif:contains

define input:inference 'http://dbpedia.org/resource/inference/rules/yago#'
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX dbpedia: <http://dbpedia.org/property/>
PREFIX yago: <http://dbpedia.org/class/yago/>

SELECT ?s ?n
FROM <http://dbpedia.org>
WHERE {
?s a yago:Publication106589574 .
?s dbpedia:name ?n .
?n bif:contains 'Lord and Rings'
}

-- Retrieve all individuals instances of the Book Class
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX dbpedia: <http://dbpedia.org/property/>
PREFIX yago: <http://dbpedia.org/class/yago/>

SELECT ?s ?n
FROM <http://dbpedia.org>
WHERE {
?s a yago:Book106410904 .
?s dbpedia:name ?n .
}

-- Retrieve all individuals instances of Publication Class which should include all Books.
define input:inference 'http://dbpedia.org/resource/inference/rules/yago#'
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX dbpedia: <http://dbpedia.org/property/>
PREFIX yago: <http://dbpedia.org/class/yago/>

SELECT ?s ?n
FROM <http://dbpedia.org>
WHERE {
?s a yago:Publication106589574 .
?s dbpedia:name ?n .
}

Note: you can also move the inference pragmas to the Virtuoso Sever side i.e place the inference rules in a server instance config file, thereby negating the need to place "define input:inference 'http://dbpedia.org/resource/inference/rules/yago#'" pragmas directly in your SPARQL queries.

Related

• Mike's UMBEL: Making Linked Data Classypost
• Fred's announcement about the Yago revamp en route to UMBEL
• Expanding Data Object Domains via UMBEL
• My Prior posts about UMBEL