Survey of Identification Techniques for Use With RDF


Second Rewrite. See related writings.

The Problem

When we use text-based languages to exchange information, we need a default "semantic map" which is a connection (in the minds of everyone using the language) between terms in the language and the things (concepts, physical objects, ...) being discussed. In working with the RDF Model and semantic web systems, it is clear that many different techniques are being used or may be used to construct that semantic map between identifiers and their denotations.

This survey focusses specifically on the mapping of character strings functioning as identifiers, like noun phrases in English. Other approaches, such as identification-by-query, can be considered a subset of this approach, if you consider the query to be expressed in some language.



Compliant with Standards Track RFCs
Exp OK
In line with public expectations (no significant group will find it fundamentally weird). An indication of whether changing the RFC is feasible.
Emb Def
Contains the definition or a way to find the definition embedded inside the identifier text
Guarantees the definition cannot change

Approach Syntax Example Denotation Characteristics


Emb Def Immut Other
1. Overloading Web
HTTP (etc)URI http://.../Creator Both a (possibly non-existant)web page and some other object, which might be documented or formally defined on the page or on a page linked from the page. no Being able to talk about pages in RDF yes no possible semantic confusion from overloading

2. Dereferencing
Terms in Web Page
HTTP (etc) URI
...22-rdf-syntax-ns#type Some object, which might be documented or formally defined on the page or on a page linked from the page. no Dublin core NSs yes no primary RDF approach
3. Dereferencing
Web Pages
deref(HTTP URI) *http://.../Creator Some object which might be2 documented or formally defined on the page. yes RDF - requires RDF2 yes no  
4. Dereferencing
Text Strings
deref("some text") englishDenotation("The person who...") Some object which can be unambiguously described in English yes machine-processable semantics yes yes may be bulky; easy to misuse through laziness
5. Minting New
uuid or tag URI Some object no Ability to dereference;

Current software.

no n/a
6. New fragid HTTP URI refernce in deref() fragid scheme. Maybe: Ambush the short form for RDF ...rdf-syntax-ns#deref(type) Object defined by the XML fragment type no messy, and messes up XML/N3 compratability. Change existing RDF.
7. Abandon XML for RDF Define N3 fragid <...rdf-syntax-ns#type> Object yes RDF/XML becomes legacy

Some of these approaches can/should be sub-categorized by what is provided in triples and what is provided at the web address, but the table is already big enough.

Sandro Hawke
$Date: 2001/04/06 03:19:00 $