The Semantics of Search - Social Semantics: The Search for Meaning on the Web - page 146

Chapter 6
The Semantics of Search

The solution to any problem in AI may be found in the writings
of Wittgenstein, though the details of implementations are rather
sketchy.

R.M. Duck-Lewis (Hirst 2000)

6.1 Introduction

What kinds of information should be used in the construction of the sense of a
resource? Given our previous work, there appears to be a priori reason why we
should confine ourselves to tags when constructing the sense of a resource. Up till
now, we have been considering the sense-based semantics of a particular URI in
the form (encoding) of a term frequency distribution. However, this seems limited.
There is always the case of co-referential URIs, where a single resource is identified
by multiple URIs. Should the semantics somehow combine the distributions of
the various Web representations? If so, precisely how - and in particular if the
Web representations are in multiple encodings? If one wanted the most thorough
description of a resource, would it not make sense to define the semantics of these
representations in terms of as many representations as possible, as it is well-known
in statistical machine-learning that there's 'no data like more data,' such that simply
adding more data under the right conditions can increase the likelihood of a stable
and rich distributional semantics (Halevy et al. 2009).
Yet the intuition that simply adding more representations to the sense will
increase its effectiveness needs to be operationalized and tested. A number of dif-
ficult questions immediately appear, such as how to identify possibly co-referential
URIs for the same resource? Or to make matters worse, how to limit the kinds of
encodings that the sense will be constructed with? These questions can be answered
by attempting to fit the intuition within a well-understood experimental paradigm,
which we believe can be the well-studied paradigm of information retrieval. To
extend further, relevance feedback is the use of explicit relevance judgments from