The processing of natural language (NL) queries and the search for
semantic matches between such queries and the contents of multimedia
documents
necessitate powerful quantifiers that
adequately model quantifying expressions in NL.
In this report, we develop an axiomatic theory of fuzzy NL quantifiers
and present a model of this theory which makes its systematical
interpretation possible. The need for such a theory arises from the
fact that the meaning of a natural language query depends heavily not
only on the concepts it contains, but also on the various quantifying
expressions interrelating these concepts in the query, which are
often fuzzy in nature.
The resulting operators form a class of generic operators for
information aggregation and the fusion of gradual evaluations, which
could also prove useful in more traditional retrieval systems.
We conclude the report by sketching some applications of the theory to
our multimedia retrieval system and give an example of how it is used
for the content-based retrieval of meteorological documents.
The DFS theory of fuzzy quantification is developed in DFS: An Axiomatic Approach to Fuzzy Quantification, which should be considered the primary reference on DFS. Our earlier axioms (in the above paper) differ slightly from DFS in its current form. However, the above report might be of interest in its own right because it is more application-oriented and describes the original motivation of developing DFS (i.e., adequate processing of natural language queries to a multimedia retrieval system).
I. Glöckner, A. Knoll
Fuzzy Quantifiers for Processing Natural-Language Queries in Content-Based Multimedia Retrieval Systems
Technical Report TR97-05, Technische Fakultät, Universität Bielefeld,
1997
PostScript.gz (24 pages, 1914k)