What Will it Take to Build the Semantic Technology Industry?

I get asked this question a lot. And I’d like to get your help in answering it please.

As co-chairman of the Semantic Technology Conference, I see lots of customer organizations experimenting and adopting semantic technologies – especially ontology-driven development projects and semantic search tools - and seemingly as many start-ups and new products emerging to address their requirements. It’s an exciting time to be in this space and I’m glad to have a part to play.

But back to the question of “what will it take?” I don’t think anyone has all the answers, though it seems there’s a growing consensus about how semantics will eventually take hold:

1. A Little Semantics Goes a Long Way
I think it was Jim Hendler who first used the expression, and I find myself in stark agreement. Much of the criticism of the semantic web vision focuses on the folly of trying to boil the ocean, yet many of the successful early adopters are getting nice results by taking small incremental steps. There’s a good exchange at Dave Beckett’s blog on this point.

2. Realistic Expectations
I guess this relates to my first point, but I remain concerned about the hype and expectations that are being set around the semantic web, and now the term Web 3.0. I, as much as anyone, would love to see the semantics field explode with growth, but this market is going to be driven by customers, not vendors, and the corporate clients I see are taking a cautious approach. I think they’ll catch on eventually, but let’s not try to push them too far, too fast. 

3. We Don’t Need a Killer App
Personally I think we need to look at semantic capabilities as an increasing component of the web and computing infrastructure, as opposed than trying to identify a killer app that’s going to kickstart a buying frenzy. If a killer app emerges then that’s great, but don’t hold your breath. There’s plenty of value to be gained in the meantime. More than anything, we need to demonstrate speedy, cheap ways to get started with semantics. This will be far more useful in the long run. 

4. We Need to Get Business Mindshare
It’s so obvious that I’m almost embarrassed to say it, but the main point is that we need to improve how we’re currently demonstrating the business value of semantic technology. I see a few key ways we can improve, starting with a greater willingness to talk about the projects already taking place. Secondly, I think we can leverage existing technology trends – especially SOA and mashups – to show how semantic technology can add value to these efforts. Third, and I might risk offending some people with this, but in the short term we should be emphasizing cost savings and reduced time to deployment over and above the extra intelligence and functionality that semantics can provide. Especially for corporate customers. Semantic SOA can save hugely over conventional approaches in data integration and interface projects, and this is where most businesses are really feeling the pain right now. 

OK, so this is a short and probably incomplete list of ideas.  Feel free to chime in here or at the Semantic Technology Conference

This year’s SemTech conference in particular will have numerous discussions around the theme of how to grow the semantic technology industry, including Mills Davis Semantic Wave 2007 tutorial, and the Keynote panel on Building the Semantic Technology Industry: A Conversation with Entrepreneurs and Investors

I hope to see you there and to get your input to the conversation.

PS: Next Tuesday (April 24) is the deadline for early registration discounts on SemTech, as well as group rates at the conference hotel in San Jose.

Comments

[Sledribbole]

Snx for you job!
It has very much helped me!

[Eric Hoffer]

Often when people speak about semantic technologies, they tend to describe something web based, that the typical surfer might unknowingly experience and enjoy – something that requires that the web be populated with semantically compliant information, and navigated by mechanisms that seek and process that compliant information… something that seems years away.

There’s some chicken and egg business going on here: we need web services that are looking for this particular form of metadata in order to justify mass creation of it, and at the same time, it needs to exist in order to justify the building of apps which take advantage of it. One approach is to build semantic layers to enable data integration and conversion, but this seems targeted at near-term fixes and band aid-type solutions to legacy systems and data sets. On the other side of the equation are vendor and consultant efforts to attract enterprises to utilize proprietary tool sets or methodologies. In either case, the focus is on generating savings or revenue.

Hitting the pause button on near-term-focused business models would do a lot to enable people to better focus on how to get from here to there, without having to worry about their own performance, paying the bills, and keeping the lights on. Everyone could then go about finally loading their photo albums, updating their address books, and populating all the metadata we need - but obviously that’s not going to happen either (echoing David Hay’s related comment).

So how do we get from here to there? This may echo some of the sentiments already expressed below/above, but in sum, I think the “way”:
i - lies at the intersection of the many popular Web 2.0-type tagging, sharing, and/or social networking apps that themselves are on fire today;
ii - comes from within activities people are already engaged in;
iii - builds from the bottom up, while at the same time being top down;
iv - is found in baby steps, but in baby steps that yield fruit with each step;
and in doing so, harnesses the natural proclivities of individuals.

In short, this means subtly re-engineering user experience in activities that people engage in (going to conferences, for example), while allowing the semantics to grow from interaction (of experts) in these activities, from the bottom up. In a recent post of mine, I talked about semantic technologies being special not just because they are the building blocks of the future – but specifically because they are building blocks designed to accommodate things we haven’t yet discovered. Using these tools to build mechanisms that allow or facilitate discovery and use of what is discovered – iteratively, for further use and discovery – is how I think we get “there”.

[Paolo Bouquet]

When we discuss about semantic technologies, I think we should distinguish between two different levels:

(1) the first is using semantic technologies for improving what we currently do with other (more standard) technologies - especially with RDBMSs - in corporate, stand-alone applications

(2) the second is building the Semantic Web.

Personally, I'm afraid that the (1) alone will never fly. RDBMS-based environments are so flexible that they can be torn to do nearly anything (IF . and the IF is important - the application is under some kind of centralized control).

So I believe that the way to bootstrap the use of semantic technologies should start from the other direction, namely proving that they can enable the "global" effect that the web was able to ignite.

However, to this end, it fear that the R&D community might have taken a dangerous path, basically trying to re-formulate most of what we knew from knowledge representation and reasoning in AI into the Semantic Web language and technology. I would suggest a different strategy.

Adrian above mentioned the analogy between the bottom-up approach for developing the Web and the Semantic Web approach. I think that there is a fundamental difference between the two stories (which cann explain why the first was so successful and the second is struggling to take off). The Web was built on top of a solid infrastructure (TCP-IP) and exploited this infrastructure to establish a global space of identifiers (URLs), which can be used both for referring to a resource (in a href link) and for retrieving the resource (through http). For the Semantic Web, we completely lack this infrastructure, as referring to an entity like "New York city" is completely different from referring to a document on the Web). In such a situation, Tim Berners-Lee's idea that the Semantic Web would make KR go global simply can't work. My opinion is that we should first enable the "web of entities" (namely getting to work a public and robust infrastructure for managing identity and reference of non digital resources on the web), and only afterwards we should try to build the "web of meanings" on top. In short: Semantic Web = web of entities + web of meanings (in this order). When this is done, then the bog players (customers) will immediatelyperceive the advantages of getting their applications aligned to this huge knowledge space, and will follow.

If anyone is interested, a personal view on how the web of entities might be enabled are presented (in a very preliminary form) in:

http://sunsite.informatik.rwth-aachen.de/Publications/CEUR-WS//Vol-201/33.pdf

[Jack Ring]

Dave,

The Semantic Technology Industry will supply technology to those who build systems that leverage the notions of meaning and understanding.

The STI will grow if it supports the building of such systems and the demonstration of value of such systems.

This claim reflects the Computational Technology Industry typified by IBM, the Seven Dwarfs, DEC, Intel, MIPS, TI and others. The software crowd made the CTI somewhat useful. The Basic programming language doubled the market. Other capbilities made the CTI output useful to 10 fold, then 100 fold then 1000-fold more users.

Semantic Web will not be the breakthrough. The current DBMS's can't handle the load. Full duplex Semantic Network Data Managers must appear.

The Basic dual of OWL will have to appear. Four times at the Delphi Group summit, today, I was asked, "How do you propose to get that knowledge out of people's heads and into 'knowledge bases'?

Semantics is not an IT thing. Tying semantics to SOA, and BPM will not help semantics because too many managers are becoming aware that IT Is the problem, not part of the solution. Semantics will do better as an ALTERNATIVE to IT.

May be helpful to 'reuse' Dee Hock's four objectives for promulgation of chaordic organizations (c.f., Birth of the Chaordic Age), paraphrased for the Semantics context.

Creation of a dozen or more examples of new, successful semantics-based solutions. Ideally, these will span such diverse areas as education, government, social services, commerce, and the environment, to demonstrate that semantic concepts have universal applicability. Develop methods and resources to help both existing and new institutions through the process of reconceiving themselves to a semantic-based behavior.

Development of visual and physical models of semantic-based sysems so that people have something to examine, experiment with, and compare to existing systems. This cannot be done with UML. The models must contain the ethical and spiritual dimensions generally lacking in current models. In addition, computer simulations will need to be created to allow people to quickly see how clarity of purpose and principles allow institutions to self-organize, evolve over decades, and link in new patterns for an enduring constructive society.

Development and dissemination of an impeccable intellectual foundation for semantic-based systems. The intellectual foundation must integrate the economic, scientific, political, historical, technical, social, and philosophical rationale for such semantic-based systems and establish the common language and metaphors necessary for widespread understanding of the concepts.

Creation of a global institution, itself a successful example of a semantics-based system, for the sole purpose of developing, disseminating, and implementing semantics concepts of organization, linking individuals, institutions, and groups of all kinds committed to institutional and societal reconstruction in a vast web of shared learning. It must enable people to pursue new organizational concepts in unique ways, on any scale, at anytime, for their own reasons

[Adrian Walker]

I think that semantic web technology so far is imitating the bottom-up, infrastructure-first approach that was so succesful for the original web.

However, it's possible, for things like "make all the world's data look like one database", that we should be also taking a top down approach -- starting at the author- and user-interface.

The top is a good place to address questions of agility, end-user-authorability by nonprogrammers, understandability, auditability, and explainability.

In fact, Web 2.0, to the extent that it is well defined, occupies some of the top down, ease of authoring space.

At our company, our top down approach is set out in our paper "A Wiki for Business Rules in Open Vocabulary, Executable English" [1]. The paper describes a system called Internet Business Logic that is live, online at www.reengineeringllc.com . The site has a number of examples that you can view, run, and change, using a browser. You can also write and run your own examples.

The general approach is outlined in the abstract of our talk at the first Semantic Technology conference [2]. The idea is to get three kinds of semantics -- data, inference, and English -- to work together seamlessly in one system.

Thanks for comments on this comment.

[1] www.reengineeringllc.com/A_Wiki_for_Business_Rules_in_Open_Vocabulary_Executable_English.pdf

[2] http://www.semantic-conference.com/program/sessions/S2.html

[Ivan Handler]

From my point of view, the business need I see is in what I call "policy oriented software." I am the CIO of a large state agency. Change requests to our systems come in all the time and are not predictable. Politicians do not like to be told that the implementation of a deal they just made will be 6 months to several years.

While there are several components to what I want to do, one is that I would like to be able to attach an ontology to each module that exists in a system so that when a policy maker is considering changes, they could query the existing system to see whether or not it can accommodate the proposed changes. If not, it could grind out a preliminary set of requirements which could be used to estimate the costs and timeframe for changes. While this is not at all simple to accomplish, it is concrete.

From what I have seen, semantic technology is now in the position of being a solution in search of a problem. Killer apps are not what is needed right now as much as concrete applications that demonstrate real benefit. I believe they are coming, but in my opinion, this is one of the major obstacles in the way of semantic technology right now.

[David Hay]

I must confess that in spite of my offering a presentation on the subject, I still feel a bit like Socrates the philosopher on the subject: I am not sure I really understand it. And if I am troubled, I have to assume that the great unwashed masses are really baffled.

What is missing is a definitive example of how a website is different because it is on the semantic web. You say you don't see the "killer app" out there, but absent something as clear as spreadsheet technology (the thing that made PC's viable) or even a web site, it is hard to make our descriptions of the semantic web real.

Google is incredibly impressive at being able to find stuff. Is that what we're talking about? But the Google people don't present at our conference, and I am sure they are most protective of their technology. Will OWL allow us to write a search engine like that? But to find a document does it have to have the right XML tags embedded in it? I certainly don't plan on doing that to the articles on my web site.

Yes, I understand each of the specific points I make in my presentation. And yes, I can successfully represent the ideas in my data model in OWL.

But what does that mean? I have sort of understood parts of the answer to that question, but I don't think I have had the ultimate "ah ha!" that I'm looking for. And I'm sure our prospective markets have not either.

That's my assignment at the Semantic Technology Conference.

Dave

[John Singer]

As long as the semantic web continues to be positioned as a means of "fixing" content on the web it won't go anywhere. The application of Ontologies as a means of data management versus traditional RDBMS technology is what will drive this industry. That being said, what is missing are robust data management tools for ontologies, a business rule language for more expressive constraints (see SBVR) and an agent programming language that is declaritive and expressed in the same manner as the Ontology itself. The business driver is easy - lower cost of developing systems using the emerging new age semantic data servers.

[Ed Addison]

Semantic Technology is not an industry, its a technology. It requires much manual labor, hence massive collaboration is required (Wiki-). Examples of successful applictions will spur more development. On the other hand, if good enough applications can be built without it, it will take a long time. Jury is out.