Kingsley Uyi Idehen's Weblog

For this particular user experience chronicle, I've simply inserted the content of Sebastian Trueg's post titled: What We Did Last Summer (And the Rest of 2009) – A Look Back Onto the Nepomuk Development Year ..., directly into this post, without any additional commentary or modification.

2009 is over. Yeah, sure, trueg, we know that, it has been over for a while now! Ok, ok, I am a bit late, but still I would like to get this one out - if only for my archive. So here goes.

Virtuoso

Let’s start with the major topic of 2009 (and also the beginning of 2010): The new Nepomuk database backend: Virtuoso. Everybody who used Nepomuk had the same problems: you either used the sesame2 backend which depends on Java and steals all of your memory or you were stuck with Redland which had the worst performance and missed some SPARQL features making important parts of Nepomuk  like queries unusable. So more than a year ago I had the idea to use the one GPL’ed database server out there that supported RDF in a professional manner: OpenLink’s Virtuoso. It has all the features we need, has a very good performance, and scales up to dimensions we will probably never reach on the desktop (yeah, right, and 64k main memory will be enough forever!). So very early I started coding the necessary Soprano plugin which would talk to a locally running Virtuoso server through ODBC. But since I ran into tons of small problems (as always) and got sidetracked by other tasks I did not finish it right away. OpenLink, however, was very interested in the idea of their server being part of every KDE installation (why wouldn’t they ;)). So they not only introduced a lite-mode which makes Virtuoso suitable for the desktop but also helped in debugging all the problems that I had left. Many test runs, patches, and a Virtuoso 5.0.12 release later I could finally announce the Virtuoso integration as usable.

Then end of last year I dropped the support for sesame2 and redland. Virtuoso is now the only supported database backend. The reason is simple: Virtuoso is way more powerful than the rest - not only in terms of performance - and it is fully implemented in C(++) without any traces of Java. Maybe even more important is the integration of the full text index which makes the previously used CLucene index unnecessary. Thus, we can finally combine full text and graph queries in one SPARQL query. This results in a cleaner API and way faster return of  search results since there is no need to combine the results from several queries anymore. A direct result of that is the new Nepomuk Query API which I will discuss later.

So now the only thing I am waiting for is the first bugfix release of Virtuoso 6, i.e. 6.0.1 which will fix the bugs that make 6.0.0 fail with Nepomuk. Should be out any day now. :)

The Nepomuk Query API

Querying data in Nepomuk pre-KDE-4.4 could be done in one of two ways: 1. Use the very limited capabilities of the ResourceManager to list resources with certain properties or of a certain type; or 2. Write your own SPARQL query using ugly QString::arg replacements.

With the introduction of Virtuoso and its awesome power we can now do pretty much everything in one query. This allowed me to finally create a query API for KDE: Nepomuk::Query::Query and friends. I won’t go into much detail here since I did that before.

All in all you should remember one thing: whenever you think about writing your own SPARQL query in a KDE application - have a look at libnepomukquery. It is very likely that you can avoid the hassle of debugging a query by using the query API.

The first nice effect of the new API (apart from me using it all over the place obviously) is the new query interface in Dolphin. Internally it simply combines a bunch of Nepomuk::Query::Term objects into a Nepomuk::Query::AndTerm. All very readable and no ugly query strings.

Dolphin Search Panel in KDE SC 4.4

Shared Desktop Ontologies

An important part of the Nepomuk research project was the creation of a set of ontologies for describing desktop resources and their metadata. After the Xesam project under the umbrella of freedesktop.org had been convinced to use RDF for describing file metadata they developed their own ontology. Thanks to Evgeny (phreedom) Egorochkin and Antonie Mylka both the Xesam ontology and the Nepomuk Information Elements Ontology were already very close in design. Thus, it was relatively easy to merge the two and be left with only one ontology to support. Since then not only KDE but also Strigi and Tracker are using the Nepomuk ontologies.

At the Gran Canaria Desktop Summit I met some of the guys from Tracker and we tried to come up with a plan to create a joint project to maintain the ontologies. This got off to a rough start as nobody really felt responsible. So I simply took the initiative and released the shared-desktop-ontologies version 0.1 in November 2009. The result was a s***-load of hate-mails and bug reports due to me breaking KDE build. But in the end it was worth it. Now the package is established and other projects can start to pick it up to create data compatible to the Nepomuk system and Tracker.

Today the ontologies (and the shared-desktop-ontologies package) are maintained in the Oscaf project at Sourceforge. The situation is far from perfect but it is a good start. If you need specific properties in the ontologies or are thinking about creating one for your own application - come and join us in the bug tracker…

Timeline KIO Slave

It was at the Akonadi meeting that Will Stephenson and myself got into talking about mimicking some Zeitgeist functionality through Nepomuk. Basically it meant gathering some data when opening and when saving files. We quickly came up with a hacky patch for KIO and KFileDialog which covered most cases and allowed us to track when a file was modified and by which application. This little experiment did not leave that state though (it will, however, this year) but another one did: Zeitgeist also provides a fuse filesystem which allows to browse the files by modification dates. Well, whatever fuse can do, KIO can do as well. Introducing the timeline:/ KIO slave which gives a calendar view onto your files.

Tips And Tricks

Well, I thought I would mention the Tips And Tricks section I wrote for the techbase. It might not be a big deal but I think it contains some valuable information in case you are using Nepomuk as a developer.

Google Summer Of Code 2009

This time around I had the privilege to mentor two students in the Google Summer of Code. Alessandro Sivieri and Adam Kidder did outstanding work on Improved Virtual Folders and the Smart File Dialog.

Adam’s work lead me to some heavy improvements in the Nepomuk KIO slaves myself which I only finished this week (more details on that coming up). Alessandro continued his work on faceted file browsing in KDE and created:

Sembrowser

Alessandro is following up on his work to make faceted file browsing a reality in 2010 (and KDE SC 4.5). Since it was too late to get faceted browsing into KDE SC 4.4 he is working on Sembrowser, a stand-alone faceted file browser which will be the grounds for experiments until the code is merged into Dolphin.

Faceted Browsing in KDE with Sembrowser

Nepomuk Workshops

In 2009 I organized the first Nepomuk workshop in Freiburg, Germany. And also the second one. While I reported properly on the first one I still owe a summary for the second one. I will get around to that - sooner or later. ;)

CMake Magic

Soprano gives us a nice command line tool to create a C++ namespace from an ontology file: onto2vocabularyclass. It produces nice convenience namespaces like Soprano::Vocabulary::NAO. Nepomuk adds another tool named nepomuk-rcgen. Both were a bit clumsy to use before. Now we have nice cmake macros which make it very simple to use both.

See the techbase article on how to use the new macros.

Bangarang

Without my knowledge (imagine that!) Andrew Lake created an amazing new media player named Bangarang - a Jamaican word for noise, chaos or disorder. This player is Nepomuk-enabled in the sense that it has a media library which lets you browse your media files based on the Nepomuk data. It remembers the number of times a song or a video has been played and when it was played last. It allows to add detail such as the TV series name, season, episode number, or actors that are in the video - all through Nepomuk (I hope we will soon get tvdb integration).

Edit metadata directly in Bangarang

Dolphin showing TV episode metadata created by Bangarang

And of course searching for it works, too...

And it is pretty, too...

I am especially excited about this since finally applications not written or mentored by me start contributing Nepomuk data.

Gran Canaria Desktop Summit

2009 was also the year of the first Gnome-KDE joint-conference. Let me make a bulletin for completeness and refer to my previous blog post reporting on my experiences on the island.

Well, that was by far not all I did in 2009 but I think I covered most of the important topics. And after all it is ‘just a blog entry’ - there is no need for completeness. Thanks for reading.

Thanks to the TechCrunch post titled: Ten Technologies That Will Rock 2010, I've been able to quickly construct a derivative post that condenses the ten item list down to a Single Technology That Will Rock 2010 :-)

Sticking with the TechCrunch layout, here is why all roads simply lead to Linked Data come 2010 and beyond:

1. The Tablet: a new form factor addition re. Internet and Web application hosts which is just another way of saying: Linked Data will be accessible from Tablet applications.
2. Geo: GPS chips are now standard features of mobile phones, so geolocation is increasingly becoming a necessary feature for any killer app. Thus, GeoSpatial Linked Data and GeopSpatial Queries are going to be a critical success factor for any endeavor that seeks to engage mobile applications developers and ultimately their end-users. Basiacally, you want to be able to perform Esoteric Search from these devices of the form: Find Vendors of a Camcorder (e.g., with a Zoom Factor: Weight Ratio of X) within a 2km Radius of my current location. Or how many items from my WishList are available from a Vendor within a 2km radius of my current location. Conversely, provide Vendors with the ability to spot potential Customers within a 2km of a given "clicks & mortar" location (e.g. BestBuy store).
3. Realtime Search: Rich Structured Profiles that leverage standards such as FOAF and FOAF+SSL will enable Highly Personalized Realtime Search (HPRS) without compromisng privacy. Tecnically, this is about WebIDs securely bound to X.509 Certificates, providing access to verifiable and highly navigable Personal Profile Data Spaces that also double as personal search index entry points.
4. Chrome OS: Just another operating system for exploiting the burgeoning Web of Linked Data
5. HTML5: Courtesy of RDFa, just another mechanism for exposing Linked Data by making HTML+RDFa a bona fide markup for metadata (i.e., format for describing real world objects via their attribute-value graphs)
6. Mobile Video: Simplifies the production and sharing of Video annotations (comments, reviews etc.) en route to creating rich Linked Discourse Data Spaces.
7. Augmented Reality: Ditto
8. Mobile Transactions: As per points 1&2 above, Vendor Discovery and Transaction Conusmation will increasingly be driven by high SDQ applications. The "Funnel Effect" (more choices based on individual preferences) will be a critical success factor for any one operating in the Mobile Transaction realm. Note, without Linked Data you cannot deliver scalable solutions that handle the combined requirements of: SDQ, "Funnel Effect", and Mobile Device form factor, will simply maginify the importance of Web accessible Linked Data.
9. Android: An additional platform for items 1-8; basically, 2010 isn't going to be an iPhone only zone. Personally, this reminds me of a battle from the past i.e., Microsoft vs Apple, re. desktop computing dominance. Google has studied history very well :-)
10. Social CRM: this is simply about applying points 1-9 alongide the construction of Linked Data from eCRM Data Spaces.

As I've stated in the past (across a variety of mediums), you cannot build applications that have long term value without addressing the following issues:

1. Data Item or Object Identity
2. Data Structure -- Data Models
3. Data Representation -- Data Model Entity & Relationships Representation mechanism (as delivered by metadata oriented markup)
4. Data Storage -- Database Management Systems
5. Data Access -- Data Access Protocols
6. Data Presentation -- How you present Views and Reports from Structured Data Sources
7. Data Security -- Data Access Policies

The items above basically showcase the very essence of the HTTP URI abstraction that drives HTTP based Linked Data; which is also the basic payload unit that underlies REST.

Conclusion

I simply hope that the next decade marks a period of broad appreciation and comprehension of Data Access, Integration, and Management issues on the parts of: application developers, integrators, analysts, end-users, and decision makers. Remember, without structured Data we cannot produce or share Information, and without Information, we cannot produce of share Knowledge.

Related

• HTTP URI Abstraction and Linked Data
• First Law of Data Quality
• Who's Data Is It?
• Serendipitous Discovery Quotient (SDQ)
• SDQ: The Future of SEO or an Abstract Concept?
• SPARQL & GeoSpatial Indexing (SPARQL-GEO)
• Mastering Your Own Search Index
• Solving the Paradox of Choice.

Thanks to the TechCrunch post titled: Ten Technologies That Will Rock 2010, I've been able to quickly construct a derivative post that condenses the ten item list down to a Single Technology That Will Rock 2010 :-)

Sticking with the TechCrunch layout, here is why all roads simply lead to Linked Data come 2010 and beyond:

1. The Tablet: a new form factor addition re. Internet and Web application hosts which is just another way of saying: Linked Data will be accessible from Tablet applications.
2. Geo: GPS chips are now standard features of mobile phones, so geolocation is increasingly becoming a necessary feature for any killer app. Thus, GeoSpatial Linked Data and GeopSpatial Queries are going to be a critical success factor for any endeavor that seeks to engage mobile applications developers and ultimately their end-users. Basiacally, you want to be able to perform Esoteric Search from these devices of the form: Find Vendors of a Camcorder (e.g., with a Zoom Factor: Weight Ratio of X) within a 2km Radius of my current location. Or how many items from my WishList are available from a Vendor within a 2km radius of my current location. Conversely, provide Vendors with the ability to spot potential Customers within a 2km of a given "clicks & mortar" location (e.g. BestBuy store).
3. Realtime Search: Rich Structured Profiles that leverage standards such as FOAF and FOAF+SSL will enable Highly Personalized Realtime Search (HPRS) without compromisng privacy. Tecnically, this is about WebIDs securely bound to X.509 Certificates, providing access to verifiable and highly navigable Personal Profile Data Spaces that also double as personal search index entry points.
4. Chrome OS: Just another operating system for exploiting the burgeoning Web of Linked Data
5. HTML5: Courtesy of RDFa, just another mechanism for exposing Linked Data by making HTML+RDFa a bona fide markup for metadata (i.e., format for describing real world objects via their attribute-value graphs)
6. Mobile Video: Simplifies the production and sharing of Video annotations (comments, reviews etc.) en route to creating rich Linked Discourse Data Spaces.
7. Augmented Reality: Ditto
8. Mobile Transactions: As per points 1&2 above, Vendor Discovery and Transaction Conusmation will increasingly be driven by high SDQ applications. The "Funnel Effect" (more choices based on individual preferences) will be a critical success factor for any one operating in the Mobile Transaction realm. Note, without Linked Data you cannot deliver scalable solutions that handle the combined requirements of: SDQ, "Funnel Effect", and Mobile Device form factor, will simply maginify the importance of Web accessible Linked Data.
9. Android: An additional platform for items 1-8; basically, 2010 isn't going to be an iPhone only zone. Personally, this reminds me of a battle from the past i.e., Microsoft vs Apple, re. desktop computing dominance. Google has studied history very well :-)
10. Social CRM: this is simply about applying points 1-9 alongide the construction of Linked Data from eCRM Data Spaces.

As I've stated in the past (across a variety of mediums), you cannot build applications that have long term value without addressing the following issues:

1. Data Item or Object Identity
2. Data Structure -- Data Models
3. Data Representation -- Data Model Entity & Relationships Representation mechanism (as delivered by metadata oriented markup)
4. Data Storage -- Database Management Systems
5. Data Access -- Data Access Protocols
6. Data Presentation -- How you present Views and Reports from Structured Data Sources
7. Data Security -- Data Access Policies

The items above basically showcase the very essence of the HTTP URI abstraction that drives HTTP based Linked Data; which is also the basic payload unit that underlies REST.

Conclusion

I simply hope that the next decade marks a period of broad appreciation and comprehension of Data Access, Integration, and Management issues on the parts of: application developers, integrators, analysts, end-users, and decision makers. Remember, without structured Data we cannot produce or share Information, and without Information, we cannot produce of share Knowledge.

Related

• HTTP URI Abstraction and Linked Data
• First Law of Data Quality
• Who's Data Is It?
• Serendipitous Discovery Quotient (SDQ)
• SDQ: The Future of SEO or an Abstract Concept?
• SPARQL & GeoSpatial Indexing (implications of SPARQL-GEO)
• Mastering Your Own Search Index
• Solving the Paradox of Choice.

Payment is a function of pain alleviation (opportunity cost) monetization.

This post is about highlighting the real pains associated with the $0.00 misconception associated with Data Access Drivers: ODBC, JDBC, ADO.NET, OLE-DB etc.

In the most basic sense, there are some fundament aspects of data access that are complex to implement and rarely implemented (if at all) by free drivers, the list includes:

1. Escape Syntaxes for Dates and Functions
2. Metadata Calls which enable smarter ODBC compliant applications (this feature is typically missing on Driver Side and abused on the Client side i.e., making clients DBMS specific by testing for specific DBMS names)
3. Scrollable Cursors, this is how you deal with change sensitivity, and most drivers actually fake support and get away with it due to shortage of applications to test proper cursor types (Static, Forward-Only, Key-Set, Dynamic, and Mixed models).

Okay, so we're done with actual driver sophistication re. implementation of critical features. Let's Up the ante by veering into the area of security. At the most basic level, It's extremely important to understand that all data access driver types provide read-write access to your databases; thus, it's imperative that data access drivers address the following:

1. Read-Only or Read-Write Access scoped to specific Users
2. Ditto applied to specific User Groups
3. Ditto applied to Database Names
4. Ditto applied to specific ODBC compliant applications
5. Ditto applied to specific ODBC host operating systems
6. Ditto applied to specific IP addresses or Ranges on your Network
7. Any combination of items 1-6 as part of a configurable data access rules/policy system.

Once you're done with security, you then have the thorny issue of data access and data flow management. In a nutshell, your driver needs to be able to handle:

1. Protection against cartesian product network flooding (e.g., user clicks on Customer Table via an ODBC compliant application without comprehension of back-end implications)
2. Enabling or Disabling of key DBMS engine data access optimization features (e.g. DBMS specific extensions exposed via Environment Variables of SQL commands based settings)
3. Conditional Connection Pooling across User, User Groups, Applications, Host Operating System, IP Address dimensions.

Once you've dealt with Security and Data Flow, you then have to address the enforcement of these settings across a myriad of ODBC compliant host, which is where Zeroconfig and centralized data access administration comes into play i.e., configure once (locally) and enforce globally.

When OpenLink Software entered the ODBC Driver Market segment in 1992, the issues above where the fundamental basis of our Multi-Tier Drivers. Thus, although we distinguished ourselves via performance, stability, and specification adherence, our fundamental engineering focus has always been skewed towards security and configurability, alongside high-performance and scalability.

As we close 2009, the security issues that pervade Native DBMS Drives, ODBC, JDBC, ADO.NET, OLE-DB etc. Drivers have only increased, courtesy of ubiquitous computing, sadly though, there remains a fundamental illusion that Data Access Drivers simply connect you to DBMS back-ends, and since you can get these drivers at $0.00 from most DBMS vendors they can't be that important.

I hope that this post brings some clarity to a very serious security and general configuration management issues associated with Data Access Drivers. Free ODBC Drivers offer nothing, when it comes to the real issues of Open Data Access. If they did, they wouldn't be worth $0.00!

Note: wondering if this has anything to do with Linked Data (my current data access focal point)? Well, remember, the Linked Data meme is fundamentally about REST based Open Data Access & Integration via HTTP; thus, what applies to Relational Model databases naturally applies to their more granular Graph Model relatives. Basically, data access security never goes away, it just gets more granular, complex, and ultimately, mercurial.

Related

• OpenLink Universal Data Access Drivers Overview - clickable diagram exposing features and benefits
• OpenLink Multi-Tier ODBC Drivers
• OpenLink Multi-Tier JDBC Drivers
• OpenLink Multi-Tier ADO.NET Providers
• Multi-Tier Drivers Overview (1993 White Paper excerpt)
• OpenLink ODBC White Paper (actual 1993 White Paper)
• OpenLink Virtuoso - which provides a Virtual Conceptual Model (via HTTP, RDF, based Linked Data) Layer above ODBC or JDBC accessible Data Sources
• Oracle Security Auditing Horror Stories -- Social Dimensions of Security compounded by Value (literal username and password) Based User Identity
• Database Security Strategies Need to Grow Up in 2010
• First Law of Data Quality
• Who's Data is it? Part 1
• Who's Data is it? Part 2

I've created a new discussion space that's squarely focused on the business development and marketing aspects of "HTTP based Linked Data" (Linked Data). As its name indicates, It's a BOLD attempt to fill a VoiD. :-)

Background

A few months ago, Aldo Bucchi posted a message to the LOD mailing list seeking a discussion space for more business and marketing oriented topic, in relation to Linked Data. At the time, my assumption was that the existing LOD mailing list served that purpose absolutely fine, but in due course I came to realize that Aldo's request had a much lager foundation than I initially suspected.

Historic Oversight

Linked Data, like its umbrella Semantic Web Project, has suffered from an inadvertent oversight on the parts of many of its enthusiasts (myself included): 100% of the discussion spaces are created by, geared towards, or dominated by researchers (from Academia primarily) and/or developers. Thus, at the very least, we've been operating in an echo chamber that only feed the existing void between the core community and those who are more interested in discussing business and marketing related topics.

The new discussion space seeks to cover the following:

1. Brainstorming Value Proposition Articulation
2. War Story Exchanges
3. Case Studies and Use-cases
4. Market Research & Positioning (for instance Linked Data is killer technology that redefines Data Integration, but none of the major research firms currently make that connection)
.

How Do I Join The Conversation? Simply sign up on the Google hosted BOLD mailing list, introduce yourself (ideally), and then start conversing! :-)

One of the real problems that pervades all routes to Linked Data value prop. incomprehension stems from the layering of its value pyramid; especially when communicating with -initially detached- end-users.

Note to Web Programmers: Linked Data is about Data (Wine) and not about Code (Fish). Thus, it isn't a "programmer only zone", far from it. More than anything else, its inherently inclusive and spreads its participation net widely across: Data Architects, Data Integrators, Power Users, Knowledge Workers, Information Workers, Data Analysts, etc.. Basically, everyone that can "click on a link" is invited to this particular party; remember, it is about "Linked Data" not "Linked Code", after all. :-)

Problematic Value Pyramid Layering

Here is an example of a Linked Data value pyramid that I am stumbling across --with some frequency-- these days (note: 1 being the pyramid apex):

1. SPARQL Queries
2. RDF Data Stores
3. RDF Data Sets
4. HTTP scheme URIs

Basically, Linked Data deployment (assigning de-referencable HTTP URIs to DBMS records, their attributes, and attribute values [optionally] ) is occurring last. Even worse, this happens in the context of Linked Open Data oriented endeavors, resulting in nothing but confusion or inadvertent perpetuation of the overarching pragmatically challenged "Semantic Web" stereotype.

As you can imagine, hitting SPARQL as your introduction to Linked Data is akin to hitting SQL as your introduction to Relational Database Technology, neither is an elevator-style value prop. relay mechanism.

In the relational realm, killer demos always started with desktop productivity tools (spreadsheets, report-writers, SQL QBE tools etc.) accessing, relational data sources en route to unveiling the "Productivity" and "Agility" value prop. that such binding delivered i.e., the desktop application (clients) and the databases (servers) are distinct, but operating in a mutually beneficial manner to all, courtesy of a data access standards such as ODBC (Open Database Connectivity).

In the Linked Data realm, learning to embrace and extend best practices from the relational dbms realm remains a challenge, a lot of this has to do with hangovers from a misguided perception that RDF databases will somehow completely replace RDBMS engines, rather than compliment them. Thus, you have a counter productive variant of NIH (Not Invented Here) in play, taking us to the dreaded realm of: Break the Pot and You Own It (exemplified by the 11+ year Semantic Web Project comprehension and appreciation odyssey).

From my vantage point, here is how I believe the Linked Data value pyramid should be layered, especially when communicating the essential value prop.:

1. HTTP URLs -- LINKs to documents (Reports) that users already appreciate, across the public Web and/or Intranets
2. HTTP URIs -- typically not visually distinguishable from the URLs, so use the Data exposed by de-referencing a URL to show how each Data Item (Entity or Object) is uniquely identified by a Generic HTTP URI, and how clicking on the said URIs leads to more structured metadata bearing documents available in a variety of data representation formats, thereby enabling flexible data presentation (e.g., smarter HTML pages)
3. SPARQL -- when a user appreciates the data representation and presentation dexterity of a Generic HTTP URI, they will be more inclined to drill down an additional layer to unravel how HTTP URIs mechanically deliver such flexibility
4. RDF Data Stores -- at this stage the user is now interested data sources behind the Generic HTTP URIs, courtesy of natural desire to tweak the data presented in the report; thus, you now have an engaged user ready to absorb the "How Generic HTTP URIs Pull This Off" message
5. RDF Data Sets -- while attempting to make or tweak HTTP URIs, users become curious about the actual data loaded into the RDF Data Store, which is where data sets used to create powerful Lookup Data Spaces (e.g., DBpedia) come into play such as those from the LOD constellation as exemplified by DBpedia (extractions from Wikipedia).

Related

• Exploring the Linked Data Value Proposition
• Simple Explanation of Linked Data & RDF Dynamics
• What is the Linked Data Meme About?
• Linked Data & Data Item Identifiers (Identity)

The recent Wikipedia imbroglio centered around DBpedia is the fundamental driver for this particular blog post. At time of writing this blog post, the DBpedia project definition in Wikipedia remains unsatisfactory due to the following shortcomings:

1. inaccurate and incomplete definition of the Project's What, Why, Who, Where, When, and How
2. inaccurate reflection of project essence, by skewing focus towards data extraction and data set dump production, which is at best a quarter of the project.

Here are some insights on DBpedia, from the perspective of someone intimately involved with the other three-quarters of the project.

What is DBpedia?

A live Web accessible RDF model database (Quad Store) derived from Wikipedia content snapshots, taken periodically. The RDF database underlies a Linked Data Space comprised of: HTML (and most recently HTML+RDFa) based data browser pages and a SPARQL endpoint.

Note: DBpedia 3.4 now exists in snapshot (warehouse) and Live Editions (currently being hot-staged). This post is about the snapshot (warehouse) edition, I'll drop a different post about the DBpedia Live Edition where a new Delta-Engine covers both extraction and database record replacement, in realtime.

When was it Created?

As an idea under the moniker "DBpedia" it was conceptualized in late 2006 by researchers at University of Leipzig (lead by Soren Auer) and Freie University, Berlin (lead by Chris Bizer). The first public instance of DBpedia (as described above) was released in February 2007. The official DBpedia coming out party occurred at WWW2007, Banff, during the inaugural Linked Data gathering, where it showcased the virtues and immense potential of TimBL's Linked Data meme.

Who's Behind It?

OpenLink Software (developers of OpenLink Virtuoso and providers of Web Hosting infrastructure), University of Leipzig, and Freie Univerity, Berlin. In addition, there is a burgeoning community of collaborators and contributors responsible DBpedia based applications, cross-linked data sets, ontologies (OpenCyc, SUMO, UMBEL, and YAGO) and other utilities. Finally, DBpedia wouldn't be possible without the global content contribution and curation efforts of Wikipedians, a point typically overlooked (albeit inadvertently).

How is it Constructed?

The steps are as follows:

1. RDF data set dump preparation via Wikipedia content extraction and transformation to RDF model data, using the N3 data representation format - Java and PHP extraction code produced and maintained by the teams at Leipzig and Berlin
2. Deployment of Linked Data that enables Data browsing and exploration using any HTTP aware user agent (e.g. basic Web Browsers) - handled by OpenLink Virtuoso (handled by Berlin via the Pubby Linked Data Server during the early months of the DBpedia project)
3. SPARQL compliant Quad Store, enabling direct access to database records via SPARQL (Query language, REST or SOAP Web Service, plus a variety of query results serialization formats) - OpenLink Virtuoso since first public release of DBpedia

In a nutshell, there are four distinct and vital components to DBpedia. Thus, DBpedia doesn't exist if all the project offered was a collection of RDF data dumps. Likewise, it doesn't exist if you have a SPARQL compliant Quad Store without loaded data sets, and of course it doesn't exist if you have a fully loaded SPARQL compliant Quad Store is up to the cocktail of challenges presented by live Web accessibility.

Why is it Important?

It remains a live exemplar for any individual or organization seeking to publishing or exploit HTTP based Linked Data on the World Wide Web. Its existence continues to stimulate growth in both density and quality of the burgeoning Web of Linked Data.

How Do I Use it?

In the most basic sense, simply browse the HTML pages en route to discovery erstwhile relationships that exist across named entities and subject matter concepts / headings. Beyond that, simply look at DBpedia as a master lookup table in a Web hosted distributed database setup; enabling you to mesh your local domain specific details with DBpedia records via structured relations (triples or 3-tuples records) comprised of HTTP URIs from both realms e.g., owl:sameAs relations.

What Can I Use it For?

Expanding on the Master-Details point above, you can use its rich URI corpus to alleviate tedium associated with activities such as:

1. List maintenance - e.g., Countries, States, Companies, Units of Measurement, Subject Headings etc.
2. Tagging - as a compliment to existing practices
3. Analytical Research - you're only a LINK (URI) away from erstwhile difficult to attain research data spread across a broad range of topics
4. Closed Vocabulary Construction - rather than commence the futile quest of building your own closed vocabulary, simply leverage Wikipedia's human curated vocabulary as our common base.

Related

• Pre-loaded and Pre-configured instances of DBpedia 3.4 - via publicly shared Amazon Elastic Block Storage Snapshots
• Virtuoso & DBpedia Tunning Guide
• What's In a Name & The Linked Data Police.

1. It isn't World Wide Web Specific (HTTP != World Wide Web)
2. It isn't Open Data Specific
3. It isn't about "Free" (Beer or Speech)
4. It isn't about Markup (so don't expect to grok it via "markup first" approach)
5. It's about Hyperdata - the use of HTTP and REST to deliver a powerful platform agnostic mechanism for Data Reference, Access, and Integration.

When trying to understand HTTP based Linked Data, especially if you're well versed in DBMS technology use (User, Power User, Architect, Analyst, DBA, or Programmer) think:

• Open Database Connectivity (ODBC) without operating system, data model, or wire-protocol specificity or lock-in potential
• Java Database Connectivity (JDBC) without programming language specificity
• ADO.NET without .NET runtime specificity and .NET bound language specificity
• OLE-DB without Windows operating system & programming language specificity
• XMLA without XML format specificity - with Tabular and Multidimensional results formats expressible in a variety of data representation formats.
• All of the above scoped to the Record rather than Container level, with Generic HTTP scheme URIs associated with each Record, Field, and Field value (optionally)

Remember the need for Data Access & Integration technology is the by product of the following realities:

1. Human curated data is ultimately dirty, because:
   • our thick thumbs, inattention, distractions, and general discomfort with typing, make typos prevalent
   • database engines exist for a variety of data models - Graph, Relational, Hierarchical;
   • within databases you have different record container/partition names e.g. Table Names;
   • within a database record container you have records that are really aspects of the same thing (different keys exist in a plethora of operational / line of business systems that expose aspects of the same entity e.g., customer data that spans Accounts, CRM, ERP application databases);
   • different field names (one database has "EMP" while another has "Employee") for the same record
   .
2. Units of measurement is driven by locale, the UK office wants to see sales in Pounds Sterling while the French office prefers Euros etc.
3. All of the above is subject to context halos which can be quite granular re. sensitivity e.g. staff travel between locations that alter locales and their roles; basically, profiles matters a lot.

Related

• ODBC and WODBC (Web Open Database Connectivity) Comparison
• Creating, Deploying, and Exploiting Linked Data Presentation
• Open Data Protocol Project

Here are 5 powerful benefits you can immediately derive from the combination of Virtuoso and Amazon's AWS services (specifically the EC2 and EBS components):

1. Acquire your own personal or service specific data space in the Cloud. Think DBase, Paradox, FoxPRO, Access of yore, but with the power of Oracle, Informix, Microsoft SQL Server etc.. using a Conceptual, as opposed to solely Logical, model based DBMS (i.e., a Hybrid DBMS Engine for: SQL, RDF, XML, and Full Text)
2. Ability to share and control access to your resources using innovations like FOAF+SSL, OpenID, and OAuth, all from one place
3. Construction of personal or organization based FOAF profiles in a matter of minutes; by simply creating a basic DBMS (or ODS application layer) account; and then using this profile to create strong links (references) to all your Data silos (esp. those from the Web 2.0 realm)
4. Load data sets from the LOD cloud or Sponge existing Web resources (i.e., on the fly data transformation to RDF model based Linked Data) and then use the combination to build powerful lookup services that enrich the value of URLs (think: Web addressable reports holding query results) that you publish
5. Bind all of the above to a domain that you own (e.g. a .Name domain) so that you have an attribution-friendly "authority" component for resource URLs and Entity URIs published from your Personal Linked Data Space on the Web (or private HTTP network).

In a nutshell, the AWS Cloud infrastructure simplifies the process of generating Federated presence on the Internet and/or World Wide Web. Remember, centralized networking models always end up creating data silos, in some context, ultimately! :-)

We have just released an Amazon EC2 based public Snapshot of DBpedia 3.4. Thus, you can now instantiate a personal and/or service specific variant of the DBpedia 3.4 Linked Data Space. Basically, you can replicate what we host, within minutes (as opposed to days). In addition, you no longer need to squabble --on an unpredictable basis with others-- for the infrastructure resources behind DBpedia's public instance, when using the SPARQL Endpoint, Faceted Search & Find Services, or HTML Browser Pages etc.

How Does It work?

1. Instantiate a Virtuoso EC2 AMI (paid variety, which is aggressively priced at $49.99 for setup and $19.99 per month thereafter)
2. Mount the shared DBpedia 3.4 public snapshot
3. Start Virtuoso Server
4. Start exploiting the DBpedia Linked Data Space.

What Interfaces are exposed?

1. SPARQL Endpoint
2. Linked Data Viewer Pages (as you see in the public DBpedia instance)
3. Faceted Search & Find UI and Web Services (REST or SOAP)
4. All the inference rules for UMBEL, SUMO, YAGO, OpenCYC, and DBpedia-OWL data dictionaries
5. Type Correlations Between DBpedia and Freebase

Enjoy!