Tuesday, September 21, 2004
Establishing Common Knowledge of URI Meaning in RDF
I just sent a version of this to the rdf-interest mailing list at the W3C. I am posting it here to further edit it.
Social convention is one of two problems I would like to comment on here, the other is symbol-sense representation.
In my view, such social manners are most fruitfully analyzed in terms of common knowledge using epistemic logic. When a URI denotes, it does so because everyone in a group knows what it denotes, and further, everyone knows that everyone knows what it denotes, and so on. In practice, of course, this process must be bounded. But it can't be avoided. There has to be a convention for publication of symbols (URIs in this case) that establishes this common knowledge of the sense it will have for the group that uses it. And it is not impossible. This is just what has been going on for years in the process of the creation of normative standards of computer languages. It involves concession to an authority such as ISO or W3C, who develops consensus among a small vanguard of potential users and then in a process of public announcement, or Kripkian baptism, network effect marketing, and compliance certification entices large groups of human and software agents to agree on the denotation of the symbols in the language. Thus in HTML, the symbol "table" denotes a certain type of graphic grid layout structure to millions of browser and editor software packages as well as to millions of human HTML authors.
In my opinion, the RDF specification is incomplete because it needs to include a standard for the decentralization and distribution of this normative standardization process itself. In a section that was ahead of its time, but not correctly worked out, the working group had a section in the specification that would have begun this. I am referring to the section about social meaning that was struck. There were many very strong objections to this. One was that it seems to give unwarranted power to organizations or individuals who would promulgate a vocabulary expressing a world view obnoxious to others. This problem is especially acute when a URI parasitically appropriates the sense of a common natural language term that no one owns. This need could be addressed by an explicit redefine performative, which would be interpreted as legally, that is, in a socially acceptable because explicit manner, breaking in a certain context with the social conventions established for the denotation of that symbol by the originating authority.
The other problem to be solved is how to represent the denotation of a symbol in a useful, machine readable manner. This is the what that is known when common knowledge is established. This problem is far more complex that it seems. The attempt to build machine readable, usable dictionaries has defied solution for years. And it may never be achieved. Take a symbol such as John Black, denoting me, I say. What does it mean to know the denotation of that symbol? If you receive a file in the FOAF vocabulary, with my IFP mailbox, do you know me? How could I ever represent what that symbol denotes to me? For starters I would need to include a complete autobiography, copies of all my art work, etc. What does it mean to my mother, to my friends, to people on this list? There is an immeasurable difference in the quality and quantity of descriptive material associated with this symbol and known to these individuals and groups.
Never-the-less this too can be done to an extent, useful for many different purposes. Last week I began collecting such machine readable representations of symbol senses on a wiki page at http://kashori.com/wikiPim/CategoryBoundedDescriptions. Among those I have collected so far are:
ConciseBoundedDescriptions
InverseFunctionalBoundedDescriptions
ChattyBoundedDescriptions
DefinateDescriptions
DiscriminantDescriptions
OntologicalClosure
PublishedSubjectIndicators
WordnetSynsets
and now
WebProperName
This last proposal appears to address some of these issues. It includes the concept of a naming authority, with a Kripkian baptism, and yet shows how it can be distributed. It concisely collects as little or as much of the descriptive properties of the entity denoted by the name and shows how to collect them in a concise structure.
Social convention is one of two problems I would like to comment on here, the other is symbol-sense representation.
In my view, such social manners are most fruitfully analyzed in terms of common knowledge using epistemic logic. When a URI denotes, it does so because everyone in a group knows what it denotes, and further, everyone knows that everyone knows what it denotes, and so on. In practice, of course, this process must be bounded. But it can't be avoided. There has to be a convention for publication of symbols (URIs in this case) that establishes this common knowledge of the sense it will have for the group that uses it. And it is not impossible. This is just what has been going on for years in the process of the creation of normative standards of computer languages. It involves concession to an authority such as ISO or W3C, who develops consensus among a small vanguard of potential users and then in a process of public announcement, or Kripkian baptism, network effect marketing, and compliance certification entices large groups of human and software agents to agree on the denotation of the symbols in the language. Thus in HTML, the symbol "table" denotes a certain type of graphic grid layout structure to millions of browser and editor software packages as well as to millions of human HTML authors.
In my opinion, the RDF specification is incomplete because it needs to include a standard for the decentralization and distribution of this normative standardization process itself. In a section that was ahead of its time, but not correctly worked out, the working group had a section in the specification that would have begun this. I am referring to the section about social meaning that was struck. There were many very strong objections to this. One was that it seems to give unwarranted power to organizations or individuals who would promulgate a vocabulary expressing a world view obnoxious to others. This problem is especially acute when a URI parasitically appropriates the sense of a common natural language term that no one owns. This need could be addressed by an explicit redefine performative, which would be interpreted as legally, that is, in a socially acceptable because explicit manner, breaking in a certain context with the social conventions established for the denotation of that symbol by the originating authority.
The other problem to be solved is how to represent the denotation of a symbol in a useful, machine readable manner. This is the what that is known when common knowledge is established. This problem is far more complex that it seems. The attempt to build machine readable, usable dictionaries has defied solution for years. And it may never be achieved. Take a symbol such as John Black, denoting me, I say. What does it mean to know the denotation of that symbol? If you receive a file in the FOAF vocabulary, with my IFP mailbox, do you know me? How could I ever represent what that symbol denotes to me? For starters I would need to include a complete autobiography, copies of all my art work, etc. What does it mean to my mother, to my friends, to people on this list? There is an immeasurable difference in the quality and quantity of descriptive material associated with this symbol and known to these individuals and groups.
Never-the-less this too can be done to an extent, useful for many different purposes. Last week I began collecting such machine readable representations of symbol senses on a wiki page at http://kashori.com/wikiPim/CategoryBoundedDescriptions. Among those I have collected so far are:
ConciseBoundedDescriptions
InverseFunctionalBoundedDescriptions
ChattyBoundedDescriptions
DefinateDescriptions
DiscriminantDescriptions
OntologicalClosure
PublishedSubjectIndicators
WordnetSynsets
and now
WebProperName
This last proposal appears to address some of these issues. It includes the concept of a naming authority, with a Kripkian baptism, and yet shows how it can be distributed. It concisely collects as little or as much of the descriptive properties of the entity denoted by the name and shows how to collect them in a concise structure.
Sunday, September 19, 2004
Web Proper Names
Web Proper Names: Naming Referents on the Web is a reference to an article that is very close to what I have been working on. The authors say, "The ability to understand something as being about something, as being oriented towards something else without any direct connection to it, is crucial to human intelligence. Any effort to make the Web more intelligent, for example by automating the exploitation of resources on the Web, will have to somehow reproduce the human ability to understand what things are about."
I agree completely with this. And furthermore, "This is an issue of immense practical importance: when someone searches the Web, they are looking for information about something. At present no automatic processes exist to index, organise, share, or even decide what web resources are about--all searches have to work with is text. The effort to provide machine-readable metadata through standards such as RDF and description logics as embodied in OWL are efforts to improve this situation. Although such efforts do allow a human to express what they believe a web-page is about in a standard way, they still beg the question of how to interoperably identify real-world things in such metadata."
Exactly correct, in my opinion. But the problem is not trivial, "Unfortunately, no-one from professional logicians to philosophers of consciousness have a solid idea about how we determine whether or not a thing is actually about something else."
But Jon Hanna has written an email with many objections to this idea.
What I have not seen in the what can URIs identify debate is the idea that the situation is quite different when you name an information resource by putting some information on a web server. In that case you have the authority to create a name for that information by virtue of your ownership and control of that information you place there. When you use a pre-existing english word you have no such naming authority. So you can't create a name. Using natural language words in URIs is sort of meaning-squating. You are highjacking an existing social convention and trying to put it to your own use. Part of the problem is this lack of understanding of the issue of naming authority.
I agree completely with this. And furthermore, "This is an issue of immense practical importance: when someone searches the Web, they are looking for information about something. At present no automatic processes exist to index, organise, share, or even decide what web resources are about--all searches have to work with is text. The effort to provide machine-readable metadata through standards such as RDF and description logics as embodied in OWL are efforts to improve this situation. Although such efforts do allow a human to express what they believe a web-page is about in a standard way, they still beg the question of how to interoperably identify real-world things in such metadata."
Exactly correct, in my opinion. But the problem is not trivial, "Unfortunately, no-one from professional logicians to philosophers of consciousness have a solid idea about how we determine whether or not a thing is actually about something else."
But Jon Hanna has written an email with many objections to this idea.
What I have not seen in the what can URIs identify debate is the idea that the situation is quite different when you name an information resource by putting some information on a web server. In that case you have the authority to create a name for that information by virtue of your ownership and control of that information you place there. When you use a pre-existing english word you have no such naming authority. So you can't create a name. Using natural language words in URIs is sort of meaning-squating. You are highjacking an existing social convention and trying to put it to your own use. Part of the problem is this lack of understanding of the issue of naming authority.
