Hi Mike,
[sent this email in both plaintext and html]
I can clearly see you trying to build a Roadmap as a part of strategy
planning -- so I will try to help you. Here is a very personal
viewpoint for your consideration. I have worked on some large projects
and this is distilled from those. I do not claim this to be "The Truth"
but a place to look from --- like new clothes, please try it on for
size :)
Comments and responses follow.
Thanks,
Arun Majumdar,
VivoMind Intelligence Inc,
(& Senior Consultant, Cutter Consortium)
Brenton, Mike wrote:
Why are we using Semantic Web Technologies to solve these problems?
This is a question we get all the time. Of course we all have some level
of stock answers, but then there are more and more questions. On the one
hand, some of the people asking this are simply detractors when it comes
to mew technologies they don't understand, while others are genuinely
concerned about the use of such a disruptive technology. It is these
people we are trying to reach by anticipating their questions and
presenting cogent arguments for semantic technologies. To assist with
this process, our team put their collective thoughts into all the
questions they have been asked and we've come up with a
not-so-exhaustive list of questions. We are putting these questions out
to people in the community so that we might get consensus-based,
well-structured, to-the-point answers.
Our situation is that we are using (trying to use) Semantic Web
Technology as an infrastructure for integration, in-depth analysis and
decision making over hundreds of concepts, millions to billions of
instances of concepts, and many terabytes of instance data held in our
knowledgebases. This data is harvested from SQL databases, extracted
from unstructured text documents, multimedia, etc. It is in this context
that we are being asked to justify using Semantic Web Technologies
rather than more traditional technologies.
Please feel free to group questions and responses as you see fit. We
appreciate any and all contributions and welcome input from both sides
of the question. Without further introduction here are the questions:
1) Is there a difference between Semantic Web Technologies and
Semantic Technologies and does it matter?
Yes. Semantic Technologies is the general name for the class of
technologies that address the notion of how to represent meaning (aka
Knowledge Representation(KR)), how to reason or carry on further
processing with the KR in accordance with requirements (I assume) to
enhance a value chain (I assume). Semantic Web Technologies form a
model instance of Semantic Technologies and the references can be found
on the web.
2) When/under what circumstances are Semantic Web Technologies the
best option, and why?
Using the NASA TRL (Technology Readiness Level) , I assess the TRL at
5/6 due to Oracle's incorporation of SOA and RDF and some examples I
have seen in environments. The world at large does not use SWEB
(Semantic Web). Therefore, I think that TRL 6/7 here answers your
question --- I was involved in a large TELCO that evaluated the
technology for hard real time: it is not usable nor ready compared to
Object methods and CORBA. We ended up using Conceptual Graphs as
master *templates* in a CORBA environment for *real time*. The SWEB
protocols and the triple stores exacted too much overhead to comply
with federal and state rules. CG templates came in easier and larger
"knowledge chunks" for the specific Order Management and Configuration
capability (using Oracle) -- this was done 4 years ago so its relevance
today, given that Oracle supports RDF, is now questionable. However,
coding with very fine grained knowledge, like RDF is like coding with
assembler --- I see RDF as the assembly language of Knowledge and CG's
as the equivalent of "C" language. But I am dismayed that turf wars
tend to obscure the benefits of both. My strongest recommendation is to
look at the Common Logic Standard :
ISO/IEC 24707 Information technology
http://cl.tamu.edu/docs/cl/24707-26-Apr-2006.pdf
http://cl.tamu.edu/docs/scl-latest.html
3) Where have Semantic Web Technologies shown improvement over
other technologies, and where it is thought to be able to do so in the
future?
Semantic Technologies that combine some SWEB have shown improvements in
certain areas (like the Telco examples) but not because their
interchange and reasoning capabilities are well represented in the
market or cpabailities space but because individual architects and
designers have leveraged specific know how in some major projects. I
haver yet to see a large scale 100% SWEB technology work, for example,
in EDI (Electronic Document Interchange) for Banking where perhaps
fraud detection capabilities may be optimized by such an effort. As it
is, the world is built using the"artistry" and know how of good
architects and designers that use what is useful and nothing more.
In the future, the Common Logic standard, the XMDR (Extended Meta Data
Registry), white pages and yellow pages, as well as the use of LDAP
(Lightwieght Directory Access Protocol) with SOAP, CL will likely
provide the best combination. RDF is useful for somthings. KIF is
useful for some things. DAML-S is useful for other things. OWL is
useful for certain things --- CL handles them all for interchanges.
4) Why is semantic technology better than using the relational
technologies we've used for years?
Semantic Technology *should* provide dynamic reconfiguration of
knowledge while relational technologies provide a static schema of
pre-existent knowledge. This distinction is between technologies that
process static data sets versus those that can handles streams of
volatile data and changing semantics. The problem is the Semantic
Technologies require new methods that can encapsulate evolvability of
control-knowledge to know *when* a schema should be changed or how to
integrate several relational stores for a logical view.
5) Why is semantic web technology better than using the object
oriented technologies?
Object Oriented (OO) technology is a simplistic model that is a special
case of Semantic Web Technology. Semantic web technology grew out of
OO. Consider that the paradigm shift in OO from procedural programming
was to *encapsulate* data with *methods*. This is powerful. It has
been used to extend data-encapsulation into Information encapsulation
(example, the Accessor methods in OO objects map data to information
and back). The next step is to encapsulate Knowledge into the OO.
That is where SWEB emerged.
Therefore, the question is not about "better or worse" but of "when"
and "why". So for ther why answer: OO models data, and with methods
provides information management. OO needs to manage Information so
that Knowledge level activities can be handled (example agents extend
OO by encapsulation of information and/or Knowledge - but this requires
*new* language methods just as OO did in its evolutions from structured
procedural programming).
6) What functionality is inherent in semantic web technology that
is not available with these traditional technologies?
The functionality is built-in logical reasoning which *enables* you to
rapidly develop capabilities. The functionality provided by OWL
editors, like Protege, allows you to focus visually on creating the
control-knowledge layer (this is formally the "Ontology") with which
you can use reasoners (like RACER) to produce *tactical responsiveness*
to your applications (ie. on specific example of tactical response is
when components in telephone answering systems can interoperate from
voice recognitiion to touch-key entry and coordinate with, for example,
Calendar software as well as logistics software to automatically
arrange meeting times.) An example of this is the USG Stamford
Passport Offices systems effort which was done several years ago.
7) Is there an advantage of ontologies over traditional relational
data schemas? What are the disadvantages?
Relational data schemas are restricted subsets of Ontologies. The issue
with relational data schemas is that the model is fixed. The theory
(ie, the language you use to work with the models) is also fixed and
the the rules are fixed (ie,. what you can do operationally). The
general notion of ontology permits working with arbitrary languages and
structures by providing commonality and reuse. For example, you
cannnot reuese Oracles's PL/SQL in IBMS DB without some effort of
translations etc.,.. An *shared* ontology (Like Ontac's efforts) is
designed to support standardization by the process of creation of
shared artefacts (like agreement upon terms and definitions).
8) Does this technology enhance our ability to achieve
standardization of data across heterogeneous sources?
Yes. However, the skill set required to do so is also still very high.
9) What is the level of open source software support for this
technology today, and is it reliable? What do you see in the future for
open source support?
IBM and others are committed to Open Source. The support is at TRL 5
(in my own estimate) and I believe that once either a Digital library
or major prohject using SOA is undertaken and successfully completed,
that at that point, we will truly know where theory hits the
application and the rubber meets the road! I see open source support
as *the* key way to go forward for commodity level software:
middleware, smart brokers or common logic runtime architecture.
10) Do you believe the technology will scale to massive data
proportions in data systems? What needs to happen for this to occur?
Yes. There needs to be a stronger shift in focus from Denotational
Semantics (ie. inventing the definitions and terms) which is a domain
almost totally dominated by Academics and Operational Semantics (where
how the artefacts are used and behaviors and capabilities developed) is
dominated by the low level programmers and designers. What needs to
happen is to build an Architecture practice with individuals that can
do both Academic theoretical work as well as operational work. This is
a restricted subset but the knowledge that could emerge from such a
group would be significant. One area where this is occurring is in the
IKRIS project ( http://nrrc.mitre.org/NRRC/ikris.htm
)
11) What does the expertise, i.e., human capitol, for working with
this technology look like today?
It is highly fragmented!!!
Skill Sets: (1) Logicians (2) Philosophers (3) Coders (4) Architects
(5) Hackers (6) Various Software Capable Scientists (those that can
code) (7) Academics (8) Competitive Small Businessess (seeking a new
niche but mistaking the development of "capability" with "market" ---
the two are functionally decoupled).
The MISSING SKILL IS: (1) Integrators (2) Systems Engineers (3)
Systems Architects
12) Who is involved in creating this technology?
I will not answer that question because I do not wish to contribute any
vendor bias here.
13) Who is working with this technology in Government?
I suggest speaking to MITRE (as an FFRDC launchind pad to get to the
list of folks). I know the partial list, but will not share it
publicly.
14) Who is working with this technology in industry (not as
vendors)?
Health, Telecommunications, Financial Services and Defense sectors ---
I cannot get specific but I was hired to integrate 200 companies in a
knowledge centric SOA effort (and I can send you a paper reference on
that if requested) in a larger enterprise and so cannot name due to
NDA's etc...
15) Who is working with this technology in the commercial software
market?
VivoMind Intelligence (my company is) and many, many others listed on
the various SBIRS (we do not do SBIRS - we provide capabiities using
the SWEB for programs and others at TRL 6).
16) Who is working with this technology in the academic world?
Stanford University --- I will be putting together more in connecting
the verb ontology to business processes and other things (but this is
very early) --- * I am looking for new sponsors on behalf of CSLI at
Stanford - please email me if interested: arun@xxxxxxxxxxxxxxxxx ---
here is the CSLI: http://www-csli.stanford.edu/ ; and my page on
this: http://www-csli.stanford.edu/~arunm/
17) What is the position of the EU on this technology and when do
they project that it will be ready for wide-scale deployment and use in
systems?
I recently spoke to several folks. Until intelligent-agent based
software with standards based interoperability and SOA is matured, SWEB
will lag behind. The first and most cirtical path is the SOA
architecture and its required infrastructure.
18) We've been promised great strides forward with XML and have seen
only limited successes. What makes this technology different?
Nothing.
19) What are promising approaches to getting data get into RDF and
OWL formats?
That sounds more like consulting engagement than an email capable
responses from me.
20) What are the principal objections and areas of discord
surrounding semantic web technologies?
I object to SWEB being seen as the panacea for everything --- SWEB is a
particular instance of a denotational, not operational, construct (with
the exception of the agent languages like FIPA, DAML etrc...). The
human brain has many modules, many different forms of representation
and many types and shapes of neurons. There are many different message
types, and contents --- I believe in XMDR, Common Logic as well as SOA
approaches as the first steps to achieving the middle ware layers
(which can then enable SWEB).
21) Since this is a disruptive technology, what is the risk that it
will be successful and available for the "long-haul?"
I have yet to see any "disruption". A language specification is only
becomes a "technology" through infrastructure support. Until SOA is
wedded as Semantic SOI (Services Oriented Infrastucture) then there can
be no disruption --- this will emerge as a function of a strong
middleware architecture. Hence, the risk factor has to be determined
in light of Service Oriented Architectures: and there are several
critical aspects to this (you can get that my paper: SOS or SOA: A
Rationale for a Semantic Service Oriented Infrastructure at Cutter
Consortium --- www.cutter.com -
http://www.cutter.com/content/itjournal/fulltext/advisor/2006/itj060208.html
)
22) In what way do relational databases employ semantics in their
design?
At the schema and Data Definition language level.
23) What is the position of the major search organizations, like
Google, on the use of semantic web technologies?
As Google! They currently use SOAP and XML in their own formats.
Regards,
Arun
Michael A. Brenton
Stone's Throw Technology, LLC
(540) 729-0392 (cell)
(703) 645-2388 (METS and CMS Projects)
(703) 375-6040 (DoDIIS Data Layer Project)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config: http://colab.cim3.net/mailman/listinfo/ontac-forum/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG
|
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config:
http://colab.cim3.net/mailman/listinfo/ontac-forum/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG (01)
|