Where Are All The RDF-based Semantic Web Apps?
RDF is the cornerstone of The Semantic Web, yet there still very few commercial RDF apps.
In the latest issue of Nodalities, a magazine about the Semantic Web by UK company Talis, there is an article by Talis CTO Ian Davis about the state of Semantic Web applications. Davis says that we're still in "Generation Zero" of the Semantic Web, because there are relatively few compelling apps. Specifically he notes that "there are still only a handful of applications that incorporate RDF at their heart and none of these are using the full potential of the Semantic Web." RDF is the Semantic Web's equivalent of the Web's HTML - its chief characteristic is the ability to ascribe meaning to data.
The few commercial RDF apps that Davis mentions are Twine, a beta knowledge management app and one of the few consumer Semantic Apps on the market today; Davis' own Talis (it has built a platform and apps such as for library management); and online reputation management tool Garlik. We also know of occasional RWW commenter Kingsley Idehen's company OpenLink Software, which is building some heavy duty RDF applications.
These can all be considered to be 'bottom up' Semantic Web apps. But most Semantic Apps today appear to be 'top down'.
Bottom Up vs Top Down
To understand the difference let's return to a classic RWW post by Alex Iskold, who wrote a primer on this topic called Top-Down: A New Approach to the Semantic Web. In that post, Alex Iskold explained that there are two main approaches to Semantic Apps:
1) Bottom Up - involves embedding
semantical annotations (meta-data) right into the data. This is the version that uses RDF extensively.
2) Top down -
relies on analyzing existing information; the ultimate
top-down solution would be a fully blown natural language processor, which is able to
understand text like people do. This rarely uses RDF, because it is seen as too complicated.
Plenty of Semantic Apps - But Most Sans RDF
We've noted a lot of 'top down' Semantic Apps in recent times. We profiled 10 of them back in November, including Freebase, Powerset, hakia, AdaptiveBlue, TripIt and more.
Startup conferences DEMO and TechCrunch50 both had a raft of Semantic startups vying for attention recently - for example see our review of SemantiFind. We've also seen some great advances this year from big companies. Thomson Reuters has been active with its OpenCalais initiative - its latest product SemanticProxy aims to give developers an easier way to extract semantic data from any web site. [Disclosure: OpenCalais is a RWW sponsor] And Yahoo, despite its ongoing struggles, has been showing excellent initative in Semantic Web with products like SearchMonkey - an open developer platform for search.
Talis' Paul Miller, who runs an excellent ZDNet blog about the Semantic Web, recently referred to a report by David Provost entitled On the cusp: a global review of the Semantic Web industry (PDF). Provost lists 17 apps, only a handful of which mentioned RDF.
It should be noted that Yahoo SearchMonkey does use at least a bit of RDF - RDFa and eRDF according to Provost's report. As Alex Iskold explained in his SearchMonkey analysis, these are variations of RDF:
"In 2005, Ian Davis, CTO of Semantic Web infrastructure company Talis, proposed eRDF - a form of RDF that can be embedded into HTML (compatible with HTML4). There is a simple mapping from eRDF to RDF so you can use any RDF/OWL vocabulary. But eRDF is not full RDF -- it has limitations. For example, there are no data types and there no blank nodes. Also, each page can only "talk" about itself and not about other pages.
Finally, the W3C published RDFa the latest embedding of RDF in XHTML, which has full RDF support. RDFa adds complexity in terms of implementation, but at the moment, gives the best way to embed RDF into HTML."
So, Where Are The Commercial RDF Apps?
The title of this post is a genuine question, because this author at least doesn't know the full answer. We know that Twine, Talis, Garlik and some others use RDF. But where are the other examples? In March, Alex suggested that RDF would be useful in places where interoperability is needed - for example the health community and in enterprises. Let us know in the comments if you know of examples in any of consumer, enterprise, health, or other markets.
Finally, dare we pose this controversial question: has the 'top down' Semantic Web "won" and we'll continue to see far more non-RDF apps in the commercial wild but relatively few RDF ones? Tim Berners-Lee and many other Semantic Web proponents will say a resounding 'no' to that. And ok, it will never be that black and white. But let us know your thoughts on whether RDF will become common place in the commercial world any time soon.
Related RWW posts on Semantic Web:
- 11 Things To Know About Semantic Web
- Semantic Web: Difficulties with the Classic Approach
- Top-Down: A New Approach to the Semantic Web
- Semantic Web Patterns: A Guide to Semantic Technologies
Flickr images: ianloic and Peter Morville
Share
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
Richard.
Any application that publishes data through an API *can* be "connected" to the Semantic Web by means of a simple adapter.
This is what the folks at OpenLink, at my company, and at various other places have been working on lately.
Can you say Microsoft Word is not web friendly?
Yes you can, but as soon as you put an HTTP server in front and serve a .DOC file, it is *on the Web*.
Same thing here: given the proper access to an API you can put the data in any system on the Semantic Web.
Just pointing out that being RDF-based or even RDF-oriented is probably not a requirement to being part of the Semantic Web.
Forget about the "semantic" label for a while. There's nothing too semantic about the semantic web, layer 1. It's all about data and URLs ;)
And this is not just wishful thinking.! New datasets get published and added on a weekly basis to the Linking Open Data cloud.
Duh! There aren't (enough) commercial applications for the Semantic Web because the Semantic-Web-with-capital-letters is a (poor) solution in search of a problem. Nobody needs it, or wants it, and it doesn't work so well in the first place.
That should be obvious by now, too. Clay Shirky has explained years ago why a Semantic Web in the traditional sense can't work:
* http://www.shirky.com/writings/semantic_syllogism.html
* http://shirky.com/writings/ontology_overrated.html
What does work is any of the grassroots approaches that have cropped up in the last few years: tag clouds, microformats, wikis (Semantic MediaWiki, anyone?) geocoding, etc. They solve actual problems - and well-delimited problems, at that - they do it in a simple way anyone can understand, and they don't force anyone's view of the world on anyone else.
Time to learn.
Felix, I guess Tim Berners-Lee has yet to be convinced by Clay Shirky's argument.
Aldo, thanks for the info, good to know.
@Felix
>the Semantic-Web-with-capital-letters is a (poor) solution in search of a problem
You might be right there...
Data integration is not a problem.
It is just a crazy idea that costs billions to the enterprise, and probably the oldest recurrent theme in IT.
> anyone can understand,
Yeah, we should learn a lesson or two from EJBs or Web Services about simplicity.
Heh. But I do give you a point there. The learning curve is a bit steep. But maybe for a good reason. Just wait 400 days.
Hi Richard,
you should look at the slides of Ivan Hermann. In the state of the Semantic Web there are many many examples of OWL/RDF apps.
http://www.w3.org/People/Ivan/CorePresentations/
The main problem is the sore lack of RDF data stores, a tiny drop in the ocean compared to the rump of unstructured data. Why sell an app for which so little material exists?
If a Fortune 500 company is going to do all the ontological work to create and embed structure in their data in the first place, they'd probably want to develop their own app to consume that data (or get IBM to do it for them), rather than buy one off the shelf from a startup.
TBL is an academic at heart. In the real world, unstructured data rules. If the semantic web means anything, its most important (and perhaps only) job is to make sense of the vast amount of unstructured data.
Kingsley Idehen and OpenLink are trying to make all the complexity of RDF, OWL, and that mess of Semantic enabling technologies disappear or be made less complex.
Unfortunately, they really need to be partnered with a big co to get the word and evangelism out. Meanwhile, they do real biz via middleware and server suites that do more than Oracle et al for the day to day database and web application market.
Keep going K.
I love the semantic web
I hate RDF
I have been hacking for more than 20 years
I still hate RDF
HTML, JS, CSS, I love them
I forever hate RDF
It is the format, stupid!
Richard: It might be because, contrary to what the old school may want to believe, RDF is unnecessarily complicated for today's semantic web... hell, even the comment thread right on this post has illustrated the confusion and even antagonism "Semantic Web" has generated. Microformats trump RDF as far as interoperability with HTML, usefulness of options, and ease of federating information... (Microformats make sense to use *in* a web application whereas RDF does not [for most practical purposes].)
Where are all the RDF-based semantic web apps? Well, at Swirrl we're working on one, though we're still in the early stages, having launched our first version last week.
We're aiming to lower the bar for efficient sharing and re-use of information in an organisation, to try to find the middle ground between individuals with their own copies of spreadsheets (easy, but poor for collaboration) and complex database systems (good collaboration, but big investment needed and can be inflexible in the face of change).
So we were looking for a lightweight flexible way to put some sort of a data model behind a collection of information, so it can be exchanged and combined in meaningful ways. After some early prototypes using a variety of approaches, we settled on RDF. People often tend to think of the semantic web as something complicated and academic. But the most important concepts of the semantic web are simple and familiar, at least to anyone who has ever tried to store information on a computer:
So for us, using RDF behind the scenes was the best fit for what we were trying to achieve. Our main aim is to improve collaboration amongst a group of colleagues - but of course exchanging information with the 'outside world' is important, and therefore it makes sense to use a standard way of representing data and it's structure.
Regarding bottom-up vs top-down, we fall more into the bottom-up camp. But because we use a spreadsheet/table of data type approach to the interface, we're in the familiar situation of labelling rows and columns. It's no extra effort for the user, and they don't have to see anything that 'looks like' RDF if they don't want to.
Echoing some of the earlier sentiment: Until now, Semantic Web applications have been a solution in search of a problem. There are some cool apps out there, but no killer apps ... at least not yet. I got excited when I saw the OpenCalais demo, but very few other people would get jazzed about an automatic tagging app that includes categorization and relevancy functionality.
It kind of reminds me of the hype relating to grid computing. Everyone sees the potential, but when the only apps are related to crunching financial and scientific data, it's hard to get really excited.
Ian Lamont
The Industry Standard
I've written about this a lot on my blog--I really believe that the semantic/linked data Web will take off when there are compelling consumer applications out there that offer up their data as linkable RDF. Once that's out there people will start to be able to link their own data out to the rest of the Web. Right now it's almost an academic exercise because the linkable data that's available isn't personal.
Applications which offer their data as RDF will start to be seen once applications start being built on RDF. Applications will start being built on RDF when developers realize the tremendous value of building on RDF.
For developers, here's the golden egg RDF lays: it frees you from the constraints of relational schemas. When you change and evolve your app, you no longer have to revise your schema, you simply add to it.
That's an extremely simple explanation, but if you're a developer you realize just how powerful it is. My suspicion is that the first RDF-based applications that see wide adoption
Back in 2003 I commented in the W3C RDF forum that RDF is way too complicated when used to solve real problems and the lack of standards, especially in query languages and representations was a huge roadblock.
http://markmail.org/message/w2ss7gwr73zqpyy7
I don't know now, but at that time, it seemed like every research group was developing its own query and representation variations.
Fast forward 5 years, it seems that still didn't change much and I was quite right. I decided to take a "wait-and-see" attitude towards SemWeb and I found it to be right attitude.
Basically people are looking for the holy grail in search, a machine so spectacularly smart that can answer questions like "what was the average milk consumption in Idaho in 1987?" Google can't answer that.
But for that, semantic web isn't the solution. I see way more future in new machine learning techniques. As storage and processing power gets cheaper and more abundant, we will see new solutions to tackle these problems in the future. And SemWeb won't be required.
@Jason Kolb
"When you change and evolve your app, you no longer have to revise your schema, you simply add to it."
Very well put! I agree that is one of the most important benefits of RDF and the graph based approach.
"It kind of reminds me of the hype relating to grid computing. Everyone sees the potential, but when the only apps are related to crunching financial and scientific data, it's hard to get really excited."
I'm curious what applications you do get excited about. More photo and video sharing?
Consumer applications have gotten so uninteresting, it's not surprising that the apps for more interesting technology are in the "boring" spaces of finance, science and logistics.
There are some outstanding comments here, thanks all.
karl, thanks for the link to Ivan's slides - there are some interesting apps in there.
Bill, I'm going to check out Swirrl - it sounds very interesting!
They will arrive as soon as everybody starts filling out their metadata properly and has a good taxonomy.
Posted by: Mark Bean
|
October 1, 2008 2:58 PM
Richard,
I've already sent a trackback ping from my blog post that adds to the developing discourse :-)
Mark, was that a tongue-in-cheek comment? If it wasn't, tell me this: where should I fill out my metadata? In a RDF file on my website? Sorry, but every time I see RDF code my brain starts shutting down (and I'm a programmer). As for good taxonomies, do read those Shirky articles I linked to earlier. The short version is, taxonomies don't work. The real world is too complex. To give an example of my own, even Carl Linné's classification of living beings, perhaps the oldest and most famous taxonomy, is challenged more often than not - lots of life forms simply don't fit in any one category.
So that's where all those Semantic Web apps are: on the hard drives of their manufacturers, waiting for a real-world system that would lend itself to easy, static classification. Good luck finding one.
Richard,
Thanks for your note!
W3C has started to publish a collection of Semantic Web Use Cases and Case Studies about a year ago at:
http://www.w3.org/2001/sw/sweo/public/UseCases/
The entries have not been written by W3C, but provided by the users themselves (W3C only publishes them). User companies and institutions include such names as Sun, Vodafone, UN's FAO, the Korean National Archives, Eli Lilly, and a bunch of others.
Sure, 33 applications is not a huge number but, again, the lists contains only those where the users have taken the extra step of writing down their experiences for W3C. The huge success of the Semantic Web Technologies event last May (in San Jose, CA) where announcement were made and money deals were closed shows that something is really moving...
Clearly, we still have a long way ahead. But things are moving rapidly...
Our approach @ Kreeo is different. Focusing on benefits to users and usability (leverage collective intelligence).
We don't use RDF as yet, have our own XML formats.
We don't use any open source products, created the entire platform scratch up with a semantic architecture.
I think it's clear that different people have different takes on the semantic web, and what they want to do with it (here's something the semantic web will NEVER be able to do - resolve ambiguity and differences in opinion).
Talking about the technology here - be it RDF / linguistic analysis / AI, user-generated / machine-compiled - doesn't make any sense unless you're clear about what you're trying to achieve, or at least which aspect of the problem you think this technology solves.
RDF (and it's meta model) is a format with which you can reach a a very high level of expressiveness; If you want to describe FACTUAL information's semantics in a VERY FINITE domain - you can probably eventually manage to do it using RDF. Which is not to be sniffed at.
I am under the impression that some people (Sir Timothy Bernard-Lee springs to mind) believe that RDF will play a major part in the realization of the glittering, holistic vision of the semantic web - a complete and sound representation of human knowledge gleaned from the web, fully searchable, inter-operable, etc.
I think that if this glittering vision is ever realized, RDF will not play a major part in it, and at any rate will not be a key-enabler for this vision. And here's why:
1)RDF is designed to be machine-readable but humans find it hard to work with (e.g. to create documents).
Note that trying to get a machine to generate RDF's for you is equivalent to solving the problem of Semantic descriptions (i.e. if we had it we wouldn't need RDF's).
2) We will NEVER be able to integrate RDF's automatically (without human intervention or dedicated code), unless they are either written based on the EXACT SAME ontology (meta-description), or unless we finally manage to create intelligent machines (in which case again we don't need RDF's).
Note that in case RDF's are REALLY widely adopted (as it should if it was to play a major part in the Semantic Web), then creating a single unambiguous ontology is equivalent to the original problem (of integrating different RDF's), only transfered to the meta-level.
Basically I don't think we should pin our hope for realizing the vision of the "glittering Semantic Web" on RDF. This vision typically means that loads of people add markups to their content and voila! their information is indexed, searchable and integrated with the rest of the world-wide-ontology.
So we've reduced the problem of the semantic web to the problem of automatic data integration. And I am sorry to remind you that this is a problem people have been working on for years and years and are not even close to finding a generic solution.
So essentially RDF cannot be a key enabler for the glittering vision of semantic web, since the problem of integrating (merging) all of those RDF's is not tractable
to date. If it were, then RDF, XML, text, binary - the format would simply not matter.
On the other hand it seems that RDF's can be useful if :
1) You've found a vertical field which is small enough to map yet not too small (otherwise it would not be useful), and you have lots of time on your hands, lots of money to invest, and lots and lots of patience (and you don't mind getting frustrated when the day comes and you WILL want to integrated it with another ontology...).
2) you are facing day-to-day software engineering problems
which result from dealing with "semantically rich" (and probably proprietary) data - like Swirrl does.
People have even used it to allow frequent data model changes in their standard apps - where relational schema became too cumbersome. But this use has very little to do with the "Semantic Web".
All in all - RDF is a description format. It's can be useful, but in my opinion only within a confined environment, where all the machines involved agree on what to do with the data it contains. Therefore I believe it's unlikely that it would be the basis of a Semantic Revolution.
Irrespective of whether the ultimate vision of 'semantic web' is fully realized, there are concepts from semantic web such as rdfa/microformats that are being used by search engines (searchmonkey) for improving search. Once the advantage is apparent, the desire to place semantic tags will take off.
The initial jumpstart will occur once most search engines publicly announce the acceptance of rdfa/microformats and browsers ( e.g. operator plugin firefox) can exploit the metadata for end user to consume. This may not be Web 3.0 but it is an important step for sufficient semantic data for web 3.0 to take off.
We all need to explain the immediate benefits to the marketing and business sponsors and not complicate the pitch with triple stores, huge databases and intelligent queries so on.
I pitched in this in a simple vein to CMS/WCM architects at my blog for the need to add microformats and update schemas to support rdf in my blog informationzen.org
Atul Kedar
Posted by: akedar.wordpress.com
|
October 2, 2008 11:02 AM
At groupswim.com, we use RDF for certain meta data. Our "application" database consists of mostly application data such as wiki pages, discussions, profiles, and files. In addition the "application" database does have meta data such as tags, ratings, computed scores for authority-on, etc.
But in addition, we also have a separate database capturing semantic relationships, the "semantics" database. Entities in this database point into our "application" database as well as other resources on the internet. Currently we use the "semantics" database to capture relationships like "citicorp is a bank", "bank is a financial-institution", "CEO is an abbreviation for Chief Executive Officer" and so forth. Groupswim sites can build their own meta data to support search, auto-tagging etc. We have also pre-built semantic structures from
existing sources such as word-net. Our semantic data-base if a RDF database, so I guess groupswim would qualify as a RDF-based semantic web-app.
As a community, we must look beyond the Semantic Web and its underlying XML, RDF, OWL infrastructure to achieve the required automation to reduce semantic complexity for the next generation of computing to emerge. Please take the time to absorb and reflect on the Cubicon Clean Slate Internet Architecture coming soon to the world wide community.
www.coretalk.net
Hi Richard,
I can appreciate your desire to ask the question, but I think it's too early to get a meaningful answer. The Semantic Web/Web of Data is at a comparable stage to the Web of Documents in the early 90s. At that point in time there was a browser, but not the kind of services and applications that emerged later and have become part of our daily lives. There was no Amazon, no eBay, no Google, no Wikipedia, and certainly no YouTube or Flickr or Google Apps.
The Semantic Web is at the same stage. We have a growing number of organisations and individuals exposing their data online in RDF according to the principles of Linked Data. We have a number of Semantic Web browsers that demonstrate how you can "surf" from one data source into another and on again by following the RDF "links" between items in different data sets. This works well to demonstrate a point, but it is not *the* point.
The Semantic Web is about linking and integrating data distributed across the Web (and maybe even behind the firewall if I choose that). Consequently Semantic Web browsers won't just be about displaying RDF documents from the Web, but about providing a particular view of something in the world (be that a place, person, movie, concept, whatever) assembled from numerous distributed data sources. Integration of heterogeneous data is a bigger challenge than just displaying one document after another as in a conventional Web browser. The result is that it's going to take a little time before we have applications that your friends and family want to use that have RDF under the hood.
In the meantime, perhaps we can move beyond Alex's top-down/bottom-up distinction - I don't think it stands up.
Firstly it seems contrary to how the terms are frequently used. "Top-down" to me evokes the widespread misconception of the Semantic Web as some attempt to create a single global ontology that everyone will agree on and will miraculously organise the world's information. This is not the point of the Semantic Web (here's why), but to add to the confusion you and Alex use the term "top down" (situated in the lower half of the diagram) to mean something completely different.
Secondly I think two (in as much as there is really a distinction) are actually complementary. Imagine some system that at great computational expense 'leverages some existing information' to generate some greater structure or meaning on which a consumer application can be built. Not everyone will want to use that application directly; they may want to reuse the data in their own application instead. Would Amazon be such a success without the Web Services API? Without some way to enable reuse of that data by third parties, these "top down" applications are looking a little bit vulnerable. RDF and friends provide this scope for easy reuse.
A better distinction than top-down/bottom-up is between data producers and data consumers. The "top down" applications can be primarily data producers and generate value by feeding raw materials into the Web for others to use when building their own applications and services. We don't yet understand all the patterns and architectures that will make this work, but as we learn more about it we'll see an increasing number of applications that use RDF at some stage of the process.
Oh, and +1 to Bill Roberts.
Richard, nice analysis. See my detailed comments in the post RDF/Linked Data Standards Not Good Enough for Intelligent Agents? Or Is It the Opposite?
We have studied the forms of expression for semantics for the last 10 years and we see that RDF and related forms are completely inadequate to express what is needed to be expressed. Here one has to consider human readable machine processable optimal forms of expression. These are not easy to arrive at! In order to arrive at these optimal forms it is necessary to place these in a structured mathematical framework.
ThoughtExpress.com has been using our own forms of semantic expression to run large enterprises with a vast reduction in complexity. No more programming.
Soon we are going to launch a new kind of web - it is based on our forms of semantics which includes deep social graph (facebook & twain are what we call shallow social graphs) in which deep relations can be expressed leading to services and commerce with an intelligent search guided by knowledge.
It's worth mentioning and applauding OpenLink
for making an open source edition of the Virtuoso server: http://virtuoso.openlinksw.com/wiki/main/
I haven't had any experience with it yet, but this will be one option for me to consider when evaluating the thankfully growing variety of RDF applications.
Posted by: michael.chelen.myopenid.com
|
October 6, 2008 10:31 PM
I think it is worth pointing out that the availability of data and the user interfaces to work with that data don't exist in large enough numbers.
Off the top of my head, I have not found many user interfaces that actually let me navigate a graph of information or other RDF'd data meaningfully.
Attraction/Repulsion graphs look pretty but don't actually solve my problem.
The only thing I have seen that works is IRC / Jabber bots backed with a little bit of personality and a triplestore.
Being able to map a sparql query to a "show me ?x of ?y" is certainly a good way to interact with small sets of data; it makes sense to me, it feels somewhat natural.
What I am seeing too, which makes me hopeful, is the spreading of bots to people using instant messengers - be it identi.ca or twitter, or livejournal's bots.
I can foresee that with a large number of existing sites pushing their API into the instant messaging space, and people becoming used to this style of interaction; that the semantic web (as a find-me-a-link or small snippet of information tool) has a chance of becoming far more real.
Hi Richard,
It will be interesting here to point about Brainwave platform A complete development and deployment platform with its own schemaless semantic database The database is RDF based, single data repository, with automated indexing, ACID compliant, memcached caching support, ability for snapshots, and mirroring, 128-bit Unified Namespace, the platform also supports JSON protocol, JQuery libraries, and also offers different template plugins, and implements capabilities based security system.
@ Daniel O'Connor: check the Brainwave online demo, and its /edit/ application after you logged in you will get what you want there.
Hello Richard,
At Cambridge Semantics all our applications are being built on RDF and the other W3C semantic technology specifications. If we did not have those we would have to invent our own versions of them to do the kinds of things we want to do. This would make no sense at all given the information networks and interoperability opportunities for our customers that will emerge from applications that are built on open standards.
However, just like raw XML or HTML, the elements of the semantic standards are just not something one should expose users to, even if they are essential under the hood. Just provided users with the benefits of semantics instead.
For an example of an upcoming application of what I mean, have a look at what what we are doing with Anzo for Excel. Customers understand the advantages of this kind of approach and yet it is full compliance with the semantic standards that drive it.
Kindest regards, Sean
Re: #30 "
There are a few standout apps that do provide comfortable interfaces, such as Semantic MediaWiki and Drupal with RDF module. If you like give SMW a try on Referata, a free host: http://referata.com/
Posted by: michael.chelen.myopenid.com
|
October 9, 2008 3:43 AM
Another startup plug: paggr. A combination of netvibes-like portals and smart data from anywhere on the web. Closed beta coming in a few weeks; teaser video and pre-registration site are up.
Posted by: bengee.myopenid.com
|
October 10, 2008 2:36 AM