Page 155 - DCAP310_INTRODUCTION_TO_ARTIFICIAL_INTELLIGENCE_AND_EXPERT_SYSTEMS

Unit 8: Structured Representation of Knowledge

8.1 Structured Knowledge Representation Notes

A means of representing relational knowledge as a labeled directed graph. Each vertex of the
graph represents a concept and each label represents a relation between concepts. Access and
updating procedures traverse and manipulate the graph. A semantic network is sometimes
regarded as a graphical notation for logical formulas.
Semantic or Associative Networks

Frame-based formalisms grew out of semantic (or associative) networks that were introduced
in the sixties as models for human semantic memory. The data organization of a semantic
network consists of labeled nodes representing concepts and labeled links representing relations
between concepts. Nodes can be used to represent both types and tokens. The is a link
hierarchically relates types to types or types to tokens; parts-of links relate concepts and their
parts, and in general any relation can be the label of a link. It is a token whereas garden, for
instance, is a type.

In a semantic network, each link between two nodes represents a separate proposition, e.g. the
fact that Lucy is a person is a proposition separate from the fact that Lucy weeds the garden and
the fact that the garden contains vegetables. But suppose that we want to represent relations
between more than two nodes. A solution is to allow nodes to represent situations or actions.
Each such node has outgoing links representing thematic roles (or cases) for participants in the
situation or action.

remove

actor object source

Lucy weed vg012

Frames

In order to incorporate more structure into semantic networks, frames were introduced by
Minsky (1975) as a representation formalism. A frame is basically a encapsulated fragment of a
semantic network with conceptually related nodes and links that can be addressed as a unity.
For example, one can define a frame for the remove action and the associated roles actor, object
and source.
The term frame is not only used for structures representing domain knowledge, but also for
structures representing linguistic knowledge. In Dells’ (1986) model for phonological encoding,
for example, a syllable frame may consist of an onset, a nucleus, and a coda. In the syllable best,
these roles are filled by b, e, and st, respectively. Several modifications have been proposed to
make frames so powerful that they can be used as high-level programming languages:
(1) constraints on the items allowed to fill each slot in the list; (2) recursive frames, the slots of
which may contain items that are themselves frames, (3) procedural attachment, in which
procedures are attached to roles of the frame to compute the fillers of these roles, and
(4) inheritance, which makes inferences along is-a links; this is dealt with in more detail in the
next topic.

LOVELY PROFESSIONAL UNIVERSITY 149