On Feb 3, 2009, at 3:47 PM, Ed Barkmeyer wrote:
Pat Hayes wrote:
No, really, this isn't accurate. The SWeb is about creating and publishing K.models to be used by software agents on the Web,
as distinct from software agents that are not "on the Web"
Yes, exactly. Which is very similar to hypertext which is on the Web (no scare quotes needed) and hypertext, of which there was a whole lot long before Sir Tim came along, not on the Web. Sorry if the Web gives you goosebumps, or whatever the problem is, but that's what the Semantic Web is all about.
My point was that if I construct an ontology to guide my data mining activities on my corporate databases, the Web is not involved. And therefore, that is NOT what OWL was designed for, if I understand right.
Right. Unless your corporate databases are somehow involved with the Web in some way, or you want to be able to access other ontologies or information sources elsewhere on the Web to help you do this data mining, and so on.
Whether OWL might yet be useful for the purpose is another matter.
Similarly, if I am engaged in electronic commerce with several suppliers using different 1990s standards, and I create an ontology for the business problem space and use it to do "semantic mediation" between message forms and "semantic validation" of the process (which we do), the Web is not involved. (The fact that the EDI and OAG and EDIFACT messages are sent over the Internet has nothing to do with the Web, and the ontology itself is not distributed, nor does it depend on anything but parts of SUMO.
Point taken, and I agree Internet does not necessarily imply Web.
) So SWeb technologies were not designed for that, and may or may not be appropriate, right?
(which made "Semantic Web technology" a brand new technology in 1995, utterly unlike the K-R and K-E work of the previous 20 years).
Nonsense. All the OWL development is directly based on the DL work dating from CLASSIC, ...
I see that my sarcasm was better veiled than I intended. ;-)
See, thats the trouble with being ironic when people can't hear the tone in your voice. Sorry I was bone-headed at this point.
If anything, my complaint about it is that its far too much old technology, in fact. There are brand new aspects and issues that arise, and it would be foolish to ignore them; and there is a new ambition that was entirely absent from any previous KR project.
[I like "ambition" here. ;-)]
I fully agree that real use of the Web for reuse and integration of ontologies is a new and important idea. But apart from using URIs, I don't see much SWeb technology actually being directed at dealing with the Web capabilities and concerns.
I realize that it took a lot of head-banging just to get RDF and OWL out there as a basis (and RIF is a real object lesson). So Pat and I probably agree on that aspect. And 7 years later, we are finally ready to take the next steps, which are the really Web-oriented ones.
Have I ever said any such thing? Of course not all KR is SWeb technology. But SWeb technology does exist, all the same.
The question is: What characterizes it?
The analogy to hypertext doesn't do much for me. It suggests that the SWeb is about distributed ontologies with distributed management. Clearly the idea of ontology repositories, versioning, and credibility and quality management would then be the primary SWeb concerns.
No, I don't think so. These are concerns, but not centrally; any more than credibility is a central
issue in e-commerce. And as for quality management, forget it. Nobody can do QM for the Web, its obvious, so nobody tries.
But I don't see that being the primary direction of the work that is labeled "SWeb technology". But maybe the problem is that we only finally have the basis for that effort.
Its more of an ambition than anything else, so far, but the mashups and demos that get SWeb folk most excited are exactly those that do use multiple data sources scattered around the Web to do something new, preferably involving a bit of real inference-making. That is the point of the whole project: using the Internet to allow remote access to formalized information, to enable new kinds of applications that wouldn't be possible without such access being available. Thats all. Maybe its a crazy idea, but the unpredictable and unpredicted success of the Web (not the SWeb) suggests, if only weakly, that its an ambition worth pursuing.
Im willing to listen to anyone who is actually using any knowledge model in any real domain. As Im sure you know, being on a standards-writing working group quickly removes any trace of ivory-tower attitude, rather in the way that being in a rock tumbler removes bumps from a pebble.
Well, for the survivors in any case. Those who wish to maintain their towers usually give up quickly.
I agree that AI's time has come and that standardization of formal languages for certain kinds of reasoning engines is appropriate. And I agree that sharing and reusing ontologies is a great idea, if we can figure out how. But that has little to do with the World Wide Web (semantic or otherwise).
I disagree. I think it has everything to do with the Web, and that to ignore the Web is to have ones head in the sand. Or maybe a better metaphor, it is to be standing in the road wondering what that light is that keeps getting brighter. But this is just a difference of opinion about the future of technology, and history will decide which of us is right.
I am willing to believe that future history may indeed support Pat's position. I only observe that the history of the Semantic Web to date does not reveal any effort to address any of the aspects of that oncoming lorry. We have only painted the stripes on a road that was laid 20 years ago.
My complaint is that the Web is not necessary to, and in many cases not significant to, a major part of the development and application of knowledge engineering.
ITs not necessary in some absolute sense, just as the Web isn't necessary for retailing. Still, any retailer who ignores the fact of the Web is likely to be losing out on an important resource, and to deliberately refuse to take account of the Web would be foolish.
Business opportunities are what they are. I am trying to identify the class of problems for which the Semantic Web, as distinct from other KR and KE work (whatever that distinction is), is envisaged to be the solution.
Document and service search is clearly a Web idea, and the Web is vital to it, and integrating the Web aspects into the languages and reasoning engines that have those purposes is critical. There are other distributed applications of ontologies that are designed to use the Web in important ways (software agents _integrally_ "on the Web", as distinct from agents that just use Internet technologies as a means of moving bits from a known remote source), but they are a very small part of the applications of knowledge engineering.
OK, so the SWeb is concerned with that part. I think its a fairly large part, and will likely be the dominant part, myself. But as I say, thats just a different prediction.
I completely agree that document and service search will be a major use of Semantic Web technologies, along with some other very unusual ones.
I want the SWeb work to be useful for "smart agents", on the Web and on other networks, because that is a future that I deem certain. But the agents are going to be most worried about reference ontologies, versions and contractual agreements, and we are having little success pushing the academic SWebbies in that direction. That is going to require some public commitments, and the destruction of several ivory towers. And the emerging public commitments are sponsoring organizations that, for various reasons, are choosing other K-R and K-E technologies and more conventional approaches. (Their solution is standard ontologies at best, and more commonly standard data models, and conflicting standards at that.)
I think you are the only person I know who identifies DL reasoning with document search. Seems to me that DL reasoners are used for many purposes, document search being only a very small proportion. The original motivation for (what are now called) DL systems was for checking subsumption relationships between data base schemas, and this is still what they are widely used for.
Well, that doesn't seem to be a Web application, so it isn't what OWL/DL is for, per the above. I'm sure OWL is quite useful for the purpose of checking completeness and consistency in DB schemas, but that wouldn't get in the door at W3C.
It doesn't have to. The W3C really doesn't care what OWL is used for, only that it gets used. This topic does however get in the door in the UK, where the Oxford group have used OWL reasoners to check the overall consistency of the National Health Service information rules (for example).
The position I take is that the "Web Ontology Language" was designed to support DL reasoning, because the computability was a requirement for the Semantic Web, and in particular, for making decisions in reasonable time
whilst doing a document/service search.
Not necessarily. Why do you believe this? I have never even seen any reference to this kind of use of OWL.
Most other valuable applications of DL reasoning are not "on the Web". Maybe I just read the wrong papers, or maybe "on the Web" is W3C-speak for "communicating".
I take it to mean, "can be accessed by http GET protocols, starting with a URI".
I agree that planning and decision making often needs inferential capability outside the scope of DL reasoning, as indeed do many other applications. This is one reason why the W3C is trying to standardize a SWeb rule language, but its turning out to be a bear, largely because there are too many kinds of 'rules' already out there.
Yup. The RuleML community was in the process of becoming a Hydra, and competing with two other groups with still other 'rules' concepts, when the RIF activity started. And trying to make RIF work with OWL and more general RDF ontologies is an exercise in mating songbirds with ostriches.
For the record, I wish that the SWeb had never been captured by the DL crowd in the first place, and had instead gone in a more common-logicky direction.
I think there are several reasons why it is better that the SWeb did go with DL. Pat cites a major one below -- industry support, which KIF and friends really didn't have. More than that, however, (and I have had to make this pitch, so maybe I shouldn't believe it) DL on-the-surface is a short step from E-R modeling and data modeling. That makes it accessible to a lot of the people, junior and senior, who had to buy into this to get a widely implemented standard.
In three recent projects, my colleagues and I have discovered that capturing knowledge in axiomatic form is not a mainstream activity. Software engineers, business analysts, librarians and manufacturing engineers can all understand UML versions of DL models with a little education, and junior knowledge engineers can convert them directly to OWL with a few corrections. But axiomatic formulation requires a specific skill that isn't easy to acquire.
But both the early Web ontology projects (DAML in the US and OIL in Europe) were packed with DL folk, and their world-view came to dominate. But note, this wasnt a W3C conspiracy to reject everyone else. I was simply a byproduct of the fact that the DL perspective was in fact the dominant one in the industrial sector represented by the W3C membership. If you feel that a different view should have predominated, you need to have arranged that a majority of people holding that view were most active in a majority of industrial-scale laboratories at the relevant time.
I completely agree with this. The existence of a known DL reasoning algorithm, and Pat's observation above, that DL was a companion to database technologies, goes a long way to explaining the prevailing industrial view.
On the subject of the Nokia use:
I meant only that the phones wouldn't work as they do, and so presumably wouldn't be bought in such numbers, if they didn't use SWeb technology.
What I was asking was: HOW do Nokia phones use or depend on SWeb technology? (Or is that a trade secret?)
Actually I don't know. I was told that they do, however (use RDF, in fact.)
If they have to use the Web to communicate with one another or to retrieve information vie Web transfer protocols, they do have a clear relationship to the Web. And it would be a serious mistake to misunderestimate the importance of this, IMO.
I don't consider the use of email to be use of the Web.
Agreed, I wasnt meaning to refer to email.
I agree that if they use a browser to upload or download their business information, they are using Web technologies. (And the same is true of other apps that use vanilla HTTP servers.) But I don't consider the use of HTTP to wrap Webservices to even be a "Web technology" -- its just the W3C version of DCE and CORBA, created to give a particular vendor a window into that market (and incidentally, to reach the 1997 CORBA competence level in 2007).
If the "SWeb project" is now the umbrella for all knowledge engineering (which I believe is the case)
? Isn't this exactly the opposite of what you were saying earlier in this message?
Pat's position is that some kind of distinction can be made. What I was arguing above was that I don't know what the distinction is, and I believe that "Semantic Web" has become a buzzword for a lot of AI research that has no clear relationship to the particular concerns of the Web. As Ian Bailey said, the term doesn't have an accepted definition.
Its an ambition rather than a statically defined set of technologies. Just like the Web itself, in fact.
I have watched very intelligent people run around painting their roses the funding color of the year for 30 years. They end up competing with a bunch of illiterate moneygrubbers with paper roses, because they won't agree to a criterion that would eliminate paper roses, if it might also eliminate old wood. And I believe "semantic Web" is now in that state.
There might well be some truth in that. Its a hot buzzword, along with "Web 2.0" and other journalistic burble. But don't blame the hype on the technologists: we aren't the ones writing the damn silly articles in WIred magazine.
:-) Pat isn't. But I have had to read NSF- and ATP-funding proposals, and Government contractor reports, and it is surprising how little they differ from the WIred articles in their introductions and in their value claims. And it is even more surprising how old, tired, varied and Web-irrelevant much of the technical content is.
We are seeing, and will see more, apps that speak to other remote apps in their operating environment and use the knowledge thus gained, with inferencing technologies, to make decisions about working cooperatively toward shared and diverse goals. These things have nothing to do with the Web per se
Are you sure? If they use Web protocols to communicate, they do have something to do with it. Seems to me that to ignore this or belittle it is a bit like saying that because air and water are both fluids, birds can swim without paying attention to the difference.
I want to make very sure that we both understand that the difference between the OPC 1999 standard for communicating with heating systems using DCOM and the OPC 2004 standard for doing exactly the same thing using XML and Webservices is just a choice of syntax that enables the programmers to use the current Microsoft DLLs.
OK, I will take your word on this. But I also want to make sure that we all understand this "just a choice of syntax" can have far-reaching effects on how these standards and the systems based on them actually get deployed in the real world. And when there get to be semantic issues that arise, these choices rapidly get much more important to interoperability than just the choice of using XML or not. Interoperability between machine systems, as I know you know, is a very delicate and fragile property that needs to be guarded and protected with extreme care.
And the fact that IEC 61850 for communication with industrial controllers now has an XML annex behind its ASN.1 annex doesn't represent a technological advance. It is just about the impact of widespread adoption of a representation standard on programmer education, which reduces the cost of software generation, just like agreeing to use Java instead of Ada.
I guess I tend to attribute more importance to such matters than you do. Its this boring, unscientific level at which 80% of the actual problems arise. Call it sociology rather than technology, its still important.
, and they need standards to know what constructs mean ("semantics"), but they have everything to do with knowledge sharing and decision making. And we are inventing terms like SmartCars and SmartGrid and SmartBuilding to describe these technologies. They may document their XML schemas using OWL information models, but the decision technologies are rules engines -- classical k-e with modern communications and processing speeds. Their apostles do not call them "semantic Web" technologies; they have no need to.
I don't give a damn what they are being called, but part of the SWeb ambition is certainly to provide communication standards to allow such things to interoperate over the Web. OWL isn't the final word: there are SWeb rule languages being designed right now.
My point is that these are knowledge engineering apps that are AI by anyone's standards, and they are the brains in the agents in a dynamic intelligent systems. They are being built on standard dictionaries and spellings for the data elements. The reference ontology for the interfaces is in the standard -- it doesn't need to be dynamically accessed. And the reference ontology that drives the agent behavior is in the agent implementation in some form.
Well, whoever is in charge of this is a very
lucky person, to have such stable, thoroughly accepted and agreed ontologies already compiled into software. Indeed, this likely doesn't need the Web for anything special. And indeed, this isn't what the SWeb is supposed to be about. So why are we arguing?
Now, it is possible that, when the agents have access to the Internet, they will look for updates to their ontologies and to the rest of their software and download them.
So, yes, they will use Web technologies to download their software updates, just like IE and the Java Suite and Photoshop do right now. And yes, they will probably communicate using XML and Webservices on whatever the local physical networks are. What Semantic Web technology does that employ?
Look, I know that defining standards isn't rocket science. Of course a standard will be a codification of work done elsewhere by others, in large part. Still, it needs to get done, and its a lot of work, much of it unrewarding and tedious. If you have technical issues with the W3C standards work that has been done, by all means shout them from the rooftops. But if all you are moaning about is journalistic hype and terminology, and the no doubt unfortunate fact that people are attracted to the latest intellectual fashion, please stop whining.
As I recall, my original contribution to this thread was an attempt to defend one of the few actual Semantic Web products -- OWL -- from complaints (1) that it was inadequate and inappropriate for whatever K-E application the original writer had in mind, and (2) that the Semantic Web could thus be dismissed as a misguided activity. My point was that the Semantic Web was not intended to cover every K-E application, and that there were clearly envisaged applications, consistent with what I perceived the "Semantic Web vision" to be, for which the existing products are useful. I think Pat and I agree on the first part of that.
I think I agree with all of it.
I wasn't whining about journalistic hype; I was complaining about "glittering generalities" (hype) from the supposed gurus of the Semantic Web, which includes one Pat Hayes. After three emails, I will admit that I now have no idea what the Semantic Web is about.
Its about making formalized information - ontologies, if you like - available on the Internet, and transmittable using Web transfer protocols. That's it. And the purpose of this is anything from pure intellectual curiosity to a strong intuition that making such access possible will enable a large number of new ways of using it for new purposes, just as the non-semantic Web opened up a huge array of new uses for hypertext and network-transfer technology (all of which predate the Web by years, but somehow didnt coalesce into the Web by themselves.) Its not a tightly describable technical area, more a messy kind of mash-up of bits of existing technology and new ideas that don't quite fit together properly, and we are still puzzling about why they don't and what to do about that. BUt in the meantime, it does exist and there has been enough work done implementing stuff that uses it, that people can play with it and try doing neat things. I don't see any reason why any of this should be perceived as a threat or a conspiracy.
You needn't fear that I will mis-characterize the activity in attempting to defend it in the future. I will do neither.
Edward J. Barkmeyer Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263 Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263 FAX: +1 301-975-4694
"The opinions expressed above do not reflect consensus of NIST,
and have not been reviewed by any Government authority."
IHMC (850)434 8903 or (650)494 3973
40 South Alcaniz St. (850)202 4416 office
Pensacola (850)202 4440 fax
FL 32502 (850)291 0667 mobile
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (01)