appropriate set of nodes and then creating the 
links between these various nodes. 
Establishing the nodes involves dividing the 
unrefined, implicit photointerpretation 
knowledge into meaningful chunks or units 
of knowledge. These nodes in most cases 
will be card-like or scrolling windows. They 
can be as simple as a single word or as 
complex as a textbook, depending on the 
granularity of knowledge representation. 
Indeed, it is the variety of nodes that 
hypermedia nodes can represent that makes 
them so versatile. Having a single topic for 
each node makes it easier for the author to 
know what links to construct. This process 
of knowledge decomposition or 
compartmentalization will continue until all 
of the available information is divided into 
individual definable nodes. 
Compartmentalization of knowledge can take 
place either in a top-down fashion orina 
bottom-up approach. If one is trying to 
convert an existing textbook or 
photointerpretation manual to a hypermedia 
system then it would make sense to start in a 
top-down fashion. In this case 
decomposition of knowledge can take 
advantage of the existing structure in the 
textbook or manual. On the other hand, if 
one is writing a new interpretation guide, it 
will make sense to compose it by building it 
from the bottom up. 
Once we have partitioned the specific 
photointerpretation knowledge into a set of 
nodes, it is necessary to define their 
interactions and decide how they are related 
to each other. This involves determining the 
linkage among these nodes, that is, to decide 
not only which nodes are related to others, 
but also the exact nature of the relationship. 
To do so we need to identify explicitly the 
various types of links between the nodes of 
the information base so that to capture the 
semantic relationships (associations) of the 
photointerpretation domain (Argialas, 
19893). Semantic relationships can define 
organizational or conceptual links. There are 
several types of links (relations or 
associations) that can be identified, such as is- 
a, is-a kind-of, contains, is contained-in, is 
adjacent to,refers to, consists of, implies, is- 
related-to, precludes, is more general than, 
leads to, is similar to, precedes, follows, is- 
an-example-of, is-a-simplification-of, 
supports, data, and others (Bielawski, 1990). 
Most photointerpretation textbooks/manuals 
have relative extensive, explicit or implicit, 
implications, relationships and 
cross-references among photointerpretation 
objects, elements, clues, and methods which 
can be used as the basis for identifying proper 
links. Good links should provide the 
means of organizing information within the 
hypermedia framework in patterns that may 
not be immediately discernable to students 
without the help of the navigational tools 
offered by the links. Indeed, good links could 
help to categorize photointerpretation 
concepts/techniques into semantically related 
378 
units or chunks which will be linked and 
accessed by association, much as a human 
being accesses related information. In this 
sense their definition and implementation 
within a system is critical to its success. The 
challenge in choosing nodes and links is to 
structure the photointerpretation knowledge in 
a way that supports the mental models that 
students may create when they use such a 
system. 
The variety of interpretations of links can add 
to the flexibility of the hypermedia information 
base but it may also contribute to 
a chaotic hyperspace. It is important to use 
photointerpretation related names for the links 
or destination nodes in order for students to be 
able to understand their navigation 
options. The number of links should depend 
on the content of each node. Every extra link 
is an additional burden on the student who has 
to determine whether or not to follow it so we 
should add only those links that are 
truly important and relevant. It is suggested 
that a good hypermedia design should tell the 
user why the destination for a link was an 
interesting place to jump to by relating it to the 
point of departure (Nielsen, 1990). 
When hypermedia system nodes and links are 
defined in a conceptual or organizational way, 
they are capable of forming semantic 
networks, that is a graphical representation of 
objects or concepts formed into nodes that are 
linked together in an associative way. 
Semantic networks are an expert system 
representation method and, thus, establishing 
conceptual links between hypeftmedia nodes 
provides the basis for a subsequent 
representation in an expert system 
environment. 
The above outlined guidelines for partitioning 
the photointerpretation knowledge in a set of 
nodes and links has assumed that the 
knowledge lies somewhere ready to be 
distributed into nodes and semantic links. 
This was a simplistic perspective. Indeed, 
there is a natural tendency to underestimate the 
difficulty of conceptualizing implicit 
knowledge. In place knowledge (textbooks, 
manuals, expertise) does not appear in some 
form that neatly fits abstract symbolic 
categories and explicit relations and 
associations like those used in computers 
(languages, databases, expert systems, 
hypermedia systems). Lying like an unmined 
and unrefined substance, implicit knowledge 
somehow enables the expert interpreter to 
recognize objects appearing directly on aerial 
images or to infer objects indirectly. 
Photointerpretation knowledge consist of a 
substantial number of concepts, facts, beliefs 
descriptions, relationships, dependencies, 
constraints, empirical associations, and 
procedures or decision rules for manipulating 
these descriptions and relationships in order to 
reason and draw conclusions. A 
hypermedia information base, needs to 
embody the right type, level, and amount of 
this knowledge in order to assist or teach 
students the photointerpretation approach.