Is there a named area of study which considers the
specifically process of of interpreting a sign together with the shared
knowledge needed by two agents who communicate (using signs)?
At one extreme, Agent 1 goes into a grocer's shop,
and presses buttons on Agent 2 for "3", "red", and "apples", and a simple
mechanical system delivers the fruit. Here the knowledge is all on Agent 1's
side, and includes both the semantics of "3", "red", and "apples", and knowledge
about vending machines.
At the other extreme, the two agents are people,
say an American tourist having got off the Paris RER in one of the suburbs, and
an Algerian shop keeper. In this, the American uses knowledge about common
social systems, and therefore identifies the context "shop" and so knows the
appropriateness of attempting to buy apples. On the other, the shop keeper
identifies the probable language from knowledge of a range of languages,
translates the phrasing to a probable match "Trois" "Pommes" and "Rouge"
(including allowing for different syntactical structures in each language), and
so on. Here both agents use a considerable amount of knowledge to be able to
communicate at all. (The complete sequence of "Hungarian Tourist Guide" sketches
by Monty Python can be used to extend the argument).
The reason for the question is that the semantic
web relies on symbols which are effectively decoded in advance (are the fixed
buttons in the first example or URIs in RDF). A major goal of the semantic web
is to broker communication between agents which either use common symbols or
equivalent symbols (sameAs). However, the business processes which stand behind
such operations ground the symbols in the artefacts and actions of the systems
operating those processes. Communication is reliably only if the symbols used by
both agents are grounded in the same way - I note that a number of the arguments
on this forum seem to be between two camps, one assuming that the grounding
problem is trivial, the other assuming that it is extremely difficult. Therefore
I am looking for a razor that can cut between the "ontologies as a formal
system" and "ontology term grounding" parts of the discussion, and so ensure
that both parts are solved.
I should also throw in my view is that the
ontology classes used by a business process are exactly those classes which
label the alternative routes onward from a decision process, and therefore
define the grounding of terms.
Sean Barker
Bristol