|
Perhaps the major obstacle to the development of computer programs
capable of the sophisticated processing of natural language is the
problem of representing and using the large and varied quantities of
world or domain knowledge that are, in general, required. This book
describes an attempt to circumvent this obstacle for one aspect of the
language processing problem - that of interpreting anaphors (pronouns
and other abbreviated expressions) in texts - by adopting a "shallow
processing" approach. In this approach, linguistic knowledge, about
syntax, semantics and local focusing, is exploited as heavily as
possible, in order to minimise reliance on world knowledge.
The ideas reported are implemented in a program called SPAR
(Shallow Processing Anaphor Resolver), which resolves anaphoric and
other linguistic ambiguities in simple English stories and generates
sentence-by-sentence paraphrases that show what interpretations have
been selected. Input to SPAR takes the form of semantic structures
derived automatically from single sentences. These structures are
integrated into a network-style text representation as processing
proceeds. To achieve anaphor resolution, SPAR combines and devel-
ops several existing techniques, most notably Sidner's theory of local
focusing and Wilks' "preference semantics" theory of semantics and
common sense inference.
Consideration of the need to resolve several anaphors in the same
sentence results in Sidner's framework being modified and extended
to allow focus-based processing to interact more flexibly with proces-
sing based on other types of knowledge. Wilks' treatment of common
sense inference is extended to incorporate a wider range of types of
inference without jeopardizing its uniformity and simplicity. Further,
his primitive-based formalism for word sense meanings is developed
in the interests of economy, accuracy and ease of use. |