The Semantics of Search - Social Semantics: The Search for Meaning on the Web - page 181

encoded in the unstructured hypertext and the structured Semantic Web representa-
tions that these two disparate sets of data can be used as relevance feedback for each
other. This leads to an exciting conclusion, and one that complexifies the earlier
picture of semantics considerably. If the Semantic Web is fundamentally about
extending the Web to those things outside the Web, then we have to acknowledge
that most of the current hypertext Web is already representational.
It is precisely the notion that sense - and therefore meaning as whole - is
'objective' that is crucial for our project of reconstructing meaning on the Web
via computational traces of user behavior. The Fregean notion of sense is identical
with our reconstructed notion of informational content . The content of information
is precisely what is shared between the source and the receiver as a result of
the conveyance of a particular message. By definition, this holding of content in
common which is the result of the transmission of an information-bearing message
must by definition involve at least two things: a source and a receiver. Furthermore,
if the source and receiver are considered to be human agents capable of speaking
natural language, then by the act of sharing sentences, which are just encodings
shared over written letters or acoustic waves in natural language, the two speakers
of language are sharing the content of those sentences. Since the content is possessed
by two people, and is by definition of information the same content, insofar as
subjective is defined to be that which is only possessed by a single agent and
objective is defined to be that which is possessed by more than one agent (although
not necessarily all agents), then content is objective .
The productive concepts from Web architecture map to the notion of a Fregean
sense rather easily. Sentences and terms natural in a language have both a syntactic
encoding and a semantic content (sense) that can realized over differing physical
substrates. This is equivalent to how encodings like HTML and RDF enable a
web-page author to share a particular sense with someone browsing the Web.
A sentence is a fully-fledged information-carrying message that can have multiple
realizations in the form of different utterances at different points in space and time,
just as the selfsame web-page can be sent from different spatially-located servers
at different times. The Gricean notion of a speaker's intentions then maps to the
meaningful behavior a sentence is supposed to engender, which can be thought of
as equivalent to the rendering and user behavior created as a result of interaction
with the Web (Grice 1957). Yet this mapping creates new problems: the problem of
sense disambiguation is now revealed to be much larger than previously supposed,
as it now stretches into all sorts of non-natural languages, ranging from logical
languages RDF to markup languages like HTML. Everything from messages in
computer protocols (formal languages) to paintings (iconic languages) are now just
encodings of information, and these too have senses and so possible ambiguities.
Representations are not just then 'in the head' but also present as an objective
component of sentences as the sense of names . In particular, a name in natural
language is no more than some encoding that has as its interpretation the sense
of a (possibly and usually distal!) referent. The class of proper names , long a source
of interest, is just a representation in natural language whose referent is an entity,
such that the name 'TimBL' refers to the person Tim Berners-Lee, while the larger