Thursday, May 26, 2005
Notational Standards
discussion à
The next generation knowledge tools
A note from Tim Berners-Lee:
(To the others on this list
of Ccs, I don't know whether you have been involved in discussion with Paul or
just added as Ccs. if the latter, I apologize for this unsolicited mail)
Paul,
I see that in the thread
which started with this message there are 29 messages from you alone. I have not read them all, but I return to the
first one Ccd me on to respond.
The entire discussion seems to be based on a misunderstanding of the Semantic Web. I am sorry that SW's use of the words Semantic and Ontology raise some incorrect expectations with those who have used the terms in other ways before. But that is natural language: unlike the semantic web, it is full of and ambiguity and thrives on it. One is free to reuse words in new ways, accepting that there is a cost to it.
You suggest that the semantic web is a top down approach. [1] It isn't - it is bottom up. One way of looking at it is that each date-oriented application now (like a calendar system or online banking or a photo album or a flight booking system) is built around a set of concepts (like event, transaction, photo, flight) which are well defined in that all the people and systems using the various data formats involved use them in ways now cast in the stone of precedent and software. For instance, you can argue philosophically about what a transaction means, but it won't change what Citibank sends you in OFX format one way or the other. [2]
The pain is that these systems don't connect. So we propose to convert all the data into a common format (at least at the abstract level), and then write rules so say things like that if there is a bank transaction in my checking account at a given time then there is an Calendar event for the same time. Then I can see bank statement events in my calendar viewer. [3]
It is bottom-up because no one centrally coordinated the banks and iCalendar folks in advance to be part of one ontology. Its bottom-up because I wrote that rule without working out what ofx:Transaction and iCal:Event and iCal:DTStart and ofx:ValidDate actually *mean*, I just related them in a way which works for me and people I work with. Others who trust that rule can use it if they like. No one has to make sure that the data and rules they use are consistent worldwide. But the logic is totally first order monotonic, not fuzzy or probabilistic. So I can build automated systems, and reuse data and rules in logical combinations to solve new problems. [4]
You talk in http://www.ontologystream.com about a "First School of Semantic Science" -- something of your own construction I assume -- in order to criticize it as being unreasonable. I hope you dissociate such criticism from the Semantic Web and W3C's standards for data integration across applications. [5]
I hope that this sets the record straight about what the SW is and isn't and doesn't and doesn't hope to do.
Tim Berners-Lee
[1] I am not sure why you think this. We feel that semantic science pulls away from the AI mythology when ontological models are developed and used in much the same way as natural language is used in normal social settings.
[2] The philosophical discussion is not about what bank transactions are. The non-interoperability of things in the information technology world are larger created because those who write the programs have vested interested in having non-interoperability. It is for this reason why I feel that your notion of semantic web will not achieve what you claim it will. The true discussion is what lies obscured by the use of RDF and OWL ontology as if these where universal knowledge representational languages. We are using the first-school second-school distinction so that the current activity under the banner “Semantic Web” is shown to be somewhat of a fraud. In your Scientific American article with Jim Hendler, it is suggested that the “Semantic Web” will be able to replace some of the judgments made regarding medical discussions. This suggestion was sexy and has lead to great confusion.
[3] It is important for me to note that this conversion to a common format is also the objective of CoreTalk , however it is important to not that this commonality is a structural commonality not a semantic one.
[4] Again, the issue is about the generality of the problems that can be solved with RDF and OWL, the main W3C standard for representing these common structural artifacts.
[5] I would like to not criticize the notion that data integration across applications is supported by the W3C paradigm. However, the criticism of over stepping what is the consequence of the W3C efforts can and should be criticized. We do not get machines that know each other, and we do not get reasoning computers. We get what we should have gotten a long time again, and that is commonality about very structural aspects of computer inter-operability. The development of semantic science cannot occur until the confusion is cleared up in the minds of most common people.